The Evaluation of Credit Union Non-Maturity Deposits

SEP.
2001
THE EVALUATION OF CREDIT UNION NON-MATURITY DEPOSITS
National Credit Union Administration
David M. Ellis, Ph.D., Senior Consultant

James V. Jordan, Ph.D., Vice President
TABLE OF CONTENTS
page
TABLE OF CONTENTS .................................................................................................................. I
LIST OF FIGURES ....................................................................................................................... III
LIST OF TABLES ......................................................................................................................... III
I. INTRODUCTION AND SUMMARY OF RECOMMENDATIONS ............................................................ 1

A. Summary of Recommendations...................................................................................... 2
B. Background on the History of Non-Maturity Deposit Valuation ........................................ 5
C. Overview......................................................................................................................... 7
II. THE DIFFERENT METHODS FOR VALUING NON-MATURITY DEPOSITS ........................................ 8

A. Overview of Valuation Theory Applied to NMDs............................................................ 11
B. Economic Rents as the Source of NMD Premia............................................................ 13
C. Valuing Deposit Liabilities Based on Economic Rents................................................... 15
D. Choosing the Discount Rate ......................................................................................... 19
E. Assumptions Underlying Specific Methods ................................................................... 20
F. Simple Vendor Methods................................................................................................ 23
G. Present Value Methods................................................................................................. 24
1. OTS Net Portfolio Value Model ................................................................................ 24
2. McGuire Performance Systems................................................................................ 25
3. Selvaggio, OTS and OAS......................................................................................... 26
H. Contingent Claims Methods .......................................................................................... 27
I. Conclusions .................................................................................................................. 28
III. CREDIT UNIONS VS. BANKS ................................................................................................. 30

A. Similarities and Dissimilarities....................................................................................... 30
1. Structural Differences............................................................................................... 30
2. Behavioral Differences ............................................................................................. 31
B. Implications for Valuing Non-Maturity Deposits ............................................................. 35
IV. THE EFFECTIVE MATURITY OF NON-MATURITY DEPOSITS ..................................................... 37

A. Duration and Effective Maturity ..................................................................................... 38
B. NMD Premia and Duration Estimates in Recent Research ........................................... 38
V. DATA AND ESTIMATION REQUIREMENTS OF DIFFERENT METHODS .......................................... 44

A. General Considerations ................................................................................................ 46
1. Data Considerations................................................................................................. 46
2. Estimation Considerations........................................................................................ 47
a. Length of sample period ...................................................................................... 47
b. Estimation method used ...................................................................................... 48
c. Frequency of updating parameter estimates........................................................ 49
B. Method-Specific Considerations ................................................................................... 49
1. Static Discounted Cash Flow Methods ..................................................................... 49
2. Contingent Claims.................................................................................................... 50
3. Option-Adjusted Spreads ......................................................................................... 50
C. Conclusions .................................................................................................................. 51
VI. AN ILLUSTRATIVE SIMPLE MODEL ........................................................................................ 52

A. Simple Valuation Results .............................................................................................. 54
B. Sensitivity Analysis ....................................................................................................... 56
VII. RECOMMENDATIONS .......................................................................................................... 62

A. Recommendations on Effective Maturity Assumptions.................................................. 62
B. Suggestions on How to Assess the Accuracy of the Recommended Effective
Maturities for Credit Union Deposits.............................................................................. 62
1. Modify the Call Report.............................................................................................. 63
2. Commission a Statistical Study of Credit Union Deposits ......................................... 63
3. Summary of Retention Rate Recommendations....................................................... 64
C. Recommendations on NMD Valuation Methods for Credit Unions ................................ 65
VIII. IMPLICATIONS AND GUIDELINES FOR NCUA ....................................................................... 67

A. General Questions........................................................................................................ 67
B. Questions Specific to Particular Methods...................................................................... 68
1. Simple vendor models.............................................................................................. 68
2. Present Value Methods ............................................................................................ 69
3. Contingent Claims methods ..................................................................................... 69
REFERENCES .......................................................................................................................... 72
APPENDIX ............................................................................................................................... 75
Summary of the Methods .................................................................................................. 75
Glossary of Terms ............................................................................................................... 80
Credit Union Non-Maturity Deposits ii

LIST OF FIGURES
page
Figure 1. Comparative Yields ................................................................................................. 32
Figure 2. Distribution of Account Balances at Banks and Credit Unions................................. 34
Figure 3. Distribution of Transactions at Banks and Credit Unions ......................................... 34
Figure 4. Sensitivity of Deposit Premiums to Assumed Decay Rate ....................................... 58
Figure 5. Sensitivity of Deposit Durations to Assumed Decay Rate........................................ 58
Figure 6. Yield Curves Used in the Non-Parallel Shift Sensitivity Analysis ............................. 59
Figure 7. Sensitivity of Deposit Premia to Spread between Deposit Rate and Market
Rate................................................................................................................................ 61
LIST OF TABLES
page
Table 1. Recommended Durations for NMDs ........................................................................... 3
Table 2. Comparison of Assumptions Underlying Different Valuation Methods for NEV ......... 21
Table 3. Comparison of Rates at Credit Unions and Banks.................................................... 31
Table 4. Premia Estimates for Non-Maturity Deposits ............................................................ 39
Table 5. Duration Estimates for Non-Maturity Deposits .......................................................... 42
Table 6. Recommended Durations for NMDs ......................................................................... 43
Table 7. Data Inputs and Methodologies ................................................................................ 45
Table 8. Industry-Wide Deposit Decay Rates for Thrift Deposits ............................................ 53
Table 9. Base Case Valuations and Results from +/- 300 bp Shocks ..................................... 55
Table 10. Sensitivity of Base Case Valuations to Assumed Maturity ...................................... 57
Table 11. The Effect of Non-Parallel Shifts in the Yield Curve................................................ 60
Table 12. Market Rate Processes .......................................................................................... 76
Table 13. Deposit Rate Processes ......................................................................................... 77
Table 14. Deposit Balance Processes.................................................................................... 78
Table 15. Valuation Equations ............................................................................................... 79
I. INTRODUCTION AND SUMMARY OF RECOMMENDATIONS
We have been asked by the National Credit Union Administration (NCUA) to evaluate the
available methods for valuing non-maturity deposits (NMDs) of individual credit unions. Non-
maturity deposits provided by credit unions, banks and thrifts have no set maturity. In the case
of credit unions, these include regular shares, share drafts and money market accounts.
We were asked to include the following tasks in this evaluation:
• Comprehensive review of the literature on valuing NMDs.
• Conceptual evaluation of alternative methods.
• Analysis of the material costs and benefits of alternative methods.
• Discussion of implementation issues for NCUA examiners and member credit unions.
• Recommendations for the most suitable valuation approaches to meet NCUA and
member credit union needs.
The NCUA was particularly interested in the question of what effective maturity may
reasonably be assumed for NMDs in simple methods that do not explicitly model NMD cash
flows.
In fulfilling this mandate, we have reviewed recent academic research and the work of
regulators and practitioners in the field, talked with economists from regulatory agencies
(including NCUA) and other consultants, performed statistical analysis of credit union data,
and performed valuations of non-maturity deposits using different valuation methodologies. In
some cases, our evaluation of vendor methods was limited by lack of publicly available
information. We have not evaluated software and information systems.
Our recommendations are based on four main criteria:
• Conceptual consistency with the theory of NMD valuation.
• Applicability to credit unions.
• Feasibility of implementation based on availability of data and ease of estimation and

application.
• Cost effectiveness, with particular attention to the needs and capabilities of different
sized credit unions.
Any method that is adopted for valuing NMDs should not be viewed in isolation, but rather in
the context of an overall asset and liability management framework. Due to the nature of the
assets and liabilities of credit unions, their balance sheets and income statements are
sensitive to changes in interest rates. Interest rate risk, defined as the sensitivity of balance
sheet items and earnings to changes in interest rates, is a major concern to managers of these
institutions and their regulators. By assessing the interest rate risk of both assets and
liabilities, one obtains a clearer picture of an institution’s exposure.
Many credit unions also calculate what is known as their “net economic value” (NEV); this is
simply the net difference between the economic (or market) value of their assets and the
economic (or market) value of their liabilities. The NEV approach does not value expected
new business activities of a credit union (in other words, expected growth in assets and
liabilities). Rather, it takes a snapshot view of the balance sheet at each quarter-end and
estimates its economic value at that time, and the value of that same balance sheet under
alternative interest rate scenarios. Therefore, our evaluation of NMD valuation methods is
based on their ability to provide reasonable estimates of interest rate risk and value in the
context of NEV.
A. Summary of Recommendations
The main conclusions and recommendations of our report are contained in Chapter VII, but
they are briefly summarized here.
1. The theoretical basis of NMD valuation is the ability of banks, thrifts and credit unions
to issue deposits that pay depositors less interest than the interest paid on
comparable-risk investments. To the extent that NMDs are free of credit risk, the
comparable-risk investment is a U.S. Treasury security. The present value of the
interest rate spread between the Treasury rate and the deposit rate (less deposit
acquisition and servicing costs) is the source of NMD value.
2. The appropriate discount rate for calculating NMD present value is the rate on a
comparable risk investment. Under the assumption of no credit risk, this is a Treasury
rate.
3. Valuation methods based on an “alternative cost of funds” approach tend to overstate

NMD value to the extent that the alternative funds, such as Eurodollar deposits
(LIBOR), have greater credit risk than the NMD. These methods provide accurate
Credit Union Non-Maturity Deposits 2

value estimates only if based on an appropriate assumption about the credit risk of
deposits and an investment benchmark of approximately equivalent credit risk.
4. Some of the simpler methods of including NMDs in asset-liability management systems

do not attempt to specify the actual cash flows of non-maturity deposits. These cash
flows include deposit withdrawals, deposit interest and deposit acquisition and
servicing costs. Methods that do not include these cash flows can be satisfactory (cost
effective) for small credit unions. For such methods, the choice of maturity is
essentially an assignment of interest rate risk to the deposit. Consequently, the
chosen maturity should be a duration (a measure of interest rate risk). We recommend
a range of durations based on durations estimated for NMDs in the literature we
surveyed. These recommended durations are given in Table 1. These durations are
the recommended final maturities, which are sometimes called “effective maturities,”
for NMDs for those credit unions that do not explicitly model NMD cash flows. If NMDs
are assigned to maturity buckets, we recommend using these durations as the NMD
maturities. If NMDs are treated as zero-coupon “bonds,” we recommend these
durations as the final deposit maturities.
Table 1. Recommended Durations for NMDs

————————————————————————————————————————————
Description: These durations are recommended as the final (“effective”) maturities of NMDs for credit
unions that do not explicitly model NMD cash flows. They are acceptable for assigning NMDs to maturity
buckets and for the final maturity when a NMD is treated as a zero-coupon bond.
————————————————————————————————————————————
Min Mid Max
Share Drafts (Transaction
Accounts) 2.0 2.5 2.8
Regular Shares (Passbook) 2.5 3.0 3.5
MMDAs 0.5 1.0 1.5
5. For credit unions that wish to base NMD valuation on actual (discounted) cash flows,
but not use the complex state-of-the-art methods, we provide an illustrative simple
present value model, which can produce durations comparable to those estimated in
the literature. The effective (final) maturities in this model are chosen so that the model
produces durations similar to those in Table 1. We do not recommend that any credit
union (or NCUA examiner) blindly adopt this model, or even this type of model, without
careful consideration of whether the assumptions make sense for a particular
institution.

6. In all cases of using simple “effective maturity” models (as in 4 above) or simple
discounted cash flow models (as in 5 above), we strongly recommend that the model
should be designed to produce durations similar to those estimated in the literature, as
summarized in Table 1.
7. We make no recommendations about the effective maturities in relatively fully-specified

models such as the Office of Thrift Supervision (OTS) NPVM and the most complex
models we surveyed known as “contingent claims” models. These models specify
future interest rate paths and the reaction of deposit balances and interest rates to
those paths. The effective (final) maturities, generally 15 – 30 years, specified in these
models are internally consistent with the other assumptions of the models. If credit
unions wish to use these complex models we do strongly recommend the following:
a. Model users (and examiners) should understand the assumptions on which the
model is based, including how future interest rate paths are defined and used,
how deposit balances and deposit rates are related to future interest rates, and
whether these assumptions conform reasonably well to deposit rate setting by
management of the institution and typical depositor behavior. For example, if
management does not expect to reduce the deposit rate as market rates fall,
then a model that assumes deposit rates will fall with market rates is not
appropriate. Similarly, if a particular credit union’s depositors are expected not
to react (by withdrawing funds) if the institution fails to raise deposit rates in a
rising rate environment, then the model should not be based on the assumption
that they will react.
b. Model parameters (the constants in the equations) should be estimated with

individual credit union data, or lacking that, with aggregate credit union data
adjusted to conform to the no-growth assumption in the NEV framework.
c. The discount rate in the model should be chosen based on the assumed default
risk of the deposits being valued.
8. Option-adjusted spread (OAS) models are generally not appropriate for credit unions.
Unlike bank deposits, there is no market for credit union deposits, which makes a key
input to the OAS method (the observed market value of the deposits) unavailable.
Some vendors have developed a proxy for this unavailable input; this can be

appropriate, but care must be taken to ensure that such proxies remain accurate
estimations of deposit premia.
9. There is a lack of data and research on the behavior of credit union non-maturity
deposits. NCUA should put in place some simple procedures that will lead to ongoing
estimates of the retention rate of deposits and commission a study to examine past
deposit behavior compared to that of banks and thrifts.
10. Because of the wide range in size and sophistication of the insured credit unions, we
have considered whether our recommendations should differ for small and large credit
unions. Generally, simple methods (see 4 and 5 above) are appropriate and cost-
effective for the smallest institutions, while the more complex methods (see 7 above)
are advised for the largest and most financially sophisticated.
B. Background on the History of Non-Maturity Deposit Valuation
Approaches to valuing non-maturity deposits (NMDs) have developed rapidly in the last ten
years, reflecting both increased sophistication in general asset/liability modeling, as well as
greater understanding of non-maturity deposits and their behavior. Ide (1999) shows how
deposit modeling has evolved through three distinct stages.
Initially, NMDs were assumed insensitive to changes in interest rates, and assigned the
longest maturity “bucket” in gap analysis, suggesting that they were not subject to the effects
of repricing. While this may seem naïve today, it partly reflects the days of “Regulation Q,”
when the Federal Reserve imposed maximum rates that could be paid on deposits, and when
there were few changes in deposit rates. Compounding the situation was the lack of
alternative investments for customers.
In addition to the lack of competitive pressures on deposit taking institutions to offer

competitive rates in order to attract and keep deposits, there was little demand for accurate
valuations of NMDs, since the risk-measurement of the times lacked rigor and sophistication.
In a period of relative stability in interest rates, and lack of competition from non-bank
institutions, there was little perceived need for the sophisticated risk analysis that is customary
today.
This cozy world was brought to an abrupt end by the events of the 1980s. In a short period,
deposit institutions had to react to a series of events that completely changed their world.
These included: the sharp rise in the level and volatility of interest rates, following the Federal

Reserve’s shift in monetary policy in October 1979; the growth of money market mutual funds;
the repeal of Regulation Q; and the creation of new products such as negotiated order of
withdrawal (NOW) accounts. Deposit-taking institutions suddenly found that unless they
offered competitive rates on competitive products that customers would take their deposits
elsewhere. New pricing strategies had to be developed, and banks and thrifts learned about
the extent to which their assets and liabilities were sensitive to changes in interest rates.
During this period, more sophisticated risk measurement and risk management tools were
developed for many parts of the deposit taking institution. Non-maturity deposits, however,
were still treated in the same way they always had been by most institutions. Some
developments in pricing models took place as academics and researchers started to analyze
the behavior of aggregate deposit balances in different economic and interest rate scenarios.
From these studies came some of the first efforts to model NMDs.
Just as deposit-taking institutions had grown accustomed to the high and volatile interest rate
environment of the 1980s, the world changed again. This time, it was the rapid decline in the
level and volatility of interest rates in the early 1990s that led banks and other institutions to
revisit their pricing strategy for NMD products again. Some asset/liability managers responded
by estimating so-called “beta-based” models. In these models, the estimated “beta”
represented a repricing percentage, i.e., the amount they should change their deposit rate
following a given change in market interest rates. This method reflected the institution’s option
to change the rate paid on deposits.
This period also saw the first attempts to model how customers might react to changes in
deposit rates and market rates. Some studies sought to follow samples of individual accounts
over time, tracking deposits, withdrawals, and transfers, in order to infer how average deposits
might behave. Other studies looked at total deposit balances, so that new accounts implicitly
replaced closed accounts, in this way inferring how an institution’s aggregate balances might
be expected to behave.1
In the most recent period, advances in other areas of asset pricing have been applied to
NMDs. This has led to the creation of complex mathematical equations to describe the
historical behavior of deposit rates and balances. In many of these methods, the options
1
The valuation methodology of McGuire Performance Solutions follows a combination of the two, using
both a sample of individual accounts and aggregate balance history.

embedded in deposits are explicitly modeled using option-pricing techniques.2 In other
methods, the market for core deposits that arose in the 1990s is used to infer the appropriate
spread (referred to as the option-adjusted spread) to add to the risk free rate when discounting
NMD cash flows.3 Examples of these latest developments are discussed in more detail in the
next chapter.
C. Overview
The remainder of this report is organized as follows: Chapter II considers the various methods
that have been developed for the valuation of non-maturity deposits, in terms of their
theoretical underpinnings. In Chapter III we consider the ways in which credit unions are
similar to and different from banks and thrifts. A reasonable range for assumptions concerning
the effective maturity of non-maturity deposits is discussed in Chapter IV. The practical
implications of using the various methods, in terms of how each method needs to be adapted
and used by individual credit unions and/or the NCUA are considered in Chapter V. In
Chapter VI we illustrate a simple discounted cash flow model. Chapter VII reports our
recommendations, and Chapter VIII discusses the implications for NCUA in the examination
process. The Appendix contains a technical summary and comparison of some of the
methods discussed in Chapter II.
2
For example, Jarrow and van Deventer (1998) and O’Brien (2000). Subsequently, we refer to these
methods as do their authors: “contingent claims” methods. Option pricing theory, developed initially
by Black and Scholes (1973), has been generalized to apply to any contingent claim, which is any
asset whose value is determined by the values of one or more “state variables.” The state variables
define the possible future states of the economy (“states of the world”) that are relevant to the value
of the asset. (For more about contingent claims, see Elliott and Kopp (1999), p. 2.)
3
For example, Selvaggio (1996).

II. THE DIFFERENT METHODS FOR VALUING NON-MATURITY DEPOSITS
The basic objective of any NMD valuation method is to project accurately, for a given interest
rate scenario, future deposit rates and balances. Regardless of the degree of complexity and
sophistication applied in valuing NMDs, all methods require assumptions about how the
deposits will behave in the future. If the assumptions err on the side of under-estimating the
sensitivity of deposit rates and balances to changes in market rates, the credit union will be
subject to unanticipated declines in profitability as the deposits reprice more quickly or are
withdrawn more quickly than expected. Conversely, if the assumptions err on the side of over-
estimating the rate sensitivity of deposit rates and balances, the credit union may invest assets
too short, forgoing more profitable long-term investments.
Any method that is adopted for valuing NMDs should not be viewed in isolation, but rather in
the context of an overall asset/liability management framework. Due to the nature of the
assets and liabilities of credit unions, their balance sheets and income statements are
sensitive to changes in interest rates. The degree of sensitivity of balance sheet items and
earnings to changes in interest rates is a matter of concern to the management of these
institutions and their regulators. By assessing the rate sensitivity of both assets and liabilities,
one obtains a clearer picture of an institution’s exposure to possible rate changes.
Ide (1999) points out that non-maturity deposit products contain two embedded options that
complicate their valuation: i) the financial institution holds the option to determine the interest
rate to pay customers, and when to change it; and ii) the customer holds the option to
withdraw all or part of the balance in the account at par. How these two options are
incorporated into a specific method largely determines the degree of complexity of that
method.4
A stated goal of regulators is to assess “the exposure to a bank’s underlying economic value
from movements in market interest rates;” in other words, the net difference between the
market value of assets and the market value of liabilities.5 This difference in market value is
4
The first option allows the issuing institution to make the deposit behave more or less as a floating rate
bond. The second option can be viewed as the option of the depositor to put the bond back to the
institution. Thus, a NMD can be viewed as a type of floating-rate, putable bond.
5
Board of Governors of the Federal Reserve, SR-96-13, “Interagency Guidance on Sound Practices for
Managing Interest Rate Risk”. See also FFIEC, “Supervisory Policy Statement on Investment
Securities and End-User Derivatives Activities” (63 Fed. Reg. 20191 at 20195, April 23, 1998), and
(continued...)

known by various names, but we shall use the term widely used in the credit union industry,
namely “Net Economic Value” (NEV).6 Another way of viewing NEV is that it is a fair value
assessment of the present value of expected future earnings, and by calculating NEV for
different interest rates the impact of changes in interest rates on earnings can be assessed.
By calculating a credit union’s NEV, management can assess the likely impact on net worth or
capital, and regulators can better assess the likelihood of failure due to market-related events.
When market value is not directly observable, a widely accepted estimate of the value of any
financial asset is the present value of expected future cash flows associated with that asset,
using a market interest rate for discounting. For default-free and option-free securities with
defined maturities, such as (non-callable) Treasury bonds, that is fairly straightforward: the
cash flows are known in advance with certainty, and the discount rate is readily observable in
the Treasury market. For option-free securities and deposits with credit risk, such as fixed-
term and fixed-rate certificates of deposit (CDs), (non-callable) corporate bonds and the like,
the maturity is known, the (expected) cash flows are readily calculated, and market discount
rates that include default risk premia can be observed. For NMDs, however, there are two
obstacles to this simple approach. First, as the name implies, the maturity of a specific NMD
is not known with certainty due to the option of the depositor to withdraw. Second, the timing
and the amount of cash flows of a specific NMD is uncertain due to both the depositor’s option
to withdraw and the credit union’s option to change the deposit rate. It follows that NMDs
cannot be precisely valued using traditional present value methods.
There are several ways to overcome these obstacles, but all involve making some simplifying
assumptions about the maturity of NMDs and about how the cash flows on the deposits vary
with interest rates and other variables.7 Some critics charge that because changing the
assumptions used in valuing NMDs can lead to material changes in the value of the deposits
and therefore in NEV, that the whole market value approach is fatally flawed and should not be
applied to credit unions. Instead, they argue, attention should be focused on measuring the
impact of changes in interest rates on book value of capital, earnings, or return on assets.
(...continued)
NCUA, “Investment Securities and End-User Derivatives: Interpretative Ruling and Policy Statement”
(63 Fed. Reg. 24097 at 24101, May 1, 1998).
6
Other terms for NEV include “Market Value of Portfolio Equity” (MVPE), “Economic Value of Equity”
(EVE), “Net Portfolio Value” (NPV), and “Market Value of Capital” (MVC).

Such an earnings exposure approach, however, is shortsighted for several reasons. First, it
assumes that accounting-based measures of risk are more accurate and appropriate than
market based measures. The experience of many savings and loans during the late 1980s
and early 1990s showed that market based measures of risk such as NEV (or NPV, as it is
called by the Office of Thrift Supervision) deteriorated long before the book value of capital
signaled the need for intervention by management and regulators.
Second, NMDs are not the only type of financial asset with indeterminate maturity and
uncertain cash flows. The best examples of such financial assets are lines of credit and bonds
with embedded options, including callable bonds, putable bonds and mortgage-backed
securities that include the homeowner’s option to prepay. The traditional methods of valuing,
for example, callable bonds was to assume they were either held to maturity or held to call and
to discount the cash flows at the yield to maturity or yield to call on a comparable callable
bond. More sophisticated valuation methods now widely used incorporate the changes in
bond cash flows that may occur due to the future exercise of the options in response to
interest rate changes.8 In other words, the complexity or lack of definition of cash flows need
not prevent valuation based on reasonable assumptions.
This leads to a third reason for not dismissing NEV: changes in value due to changes in the
assumptions do not invalidate the model. Rather, it means that management and examiners
must be careful that the assumptions made to value NMDs are appropriate for the institution in
the circumstances in which they are to be applied. A goal of this report, therefore, is to provide
some guidelines for the conditions under which some of the more common assumptions are
reasonable.
It is important to keep in mind that any model that attempts to measure risk or value is based
upon a set of assumptions. This is as true for accounting-based earnings exposure methods
as it is for economic value exposure methods such as NEV or the more complex Value-at-
Risk. One cannot question the validity of a model simply because its results are contingent
upon assumptions. All models outputs are contingent upon assumptions. For example,
accounting based measures of risk, such as book value of capital or return on assets, are
based on the assumption that historical cost measures are an accurate reflection of the value
(...continued)
7
The specific issue of the effective maturity of NMDs is considered in Chapter IV.
8
See, for example, Fabozzi (2000).

of assets. However, if a financial institution has to be taken over by a deposit insurance fund,
what matters is the market value of those assets.
A. Overview of Valuation Theory Applied to NMDs
NMDs are valued in a discounted cash flow, or present value, framework. The model
specification choices, therefore, are defining the maturity, specifying the cash flows and
choosing the discount rate. The overarching question, however, which in turn drives these
specification choices, is how the uncertainty of future cash flows is captured in the model.
In the most advanced NMD models, there is one source of uncertainty (equivalently, one “state
variable” defining the possible future “states of nature”), the market interest rate.9 In all of
these models, the market interest rate is defined as a Treasury rate. The Treasury rate is
assumed to follow a random path, or “stochastic process.” The mathematical equation of the
random path also implies the probability distribution of the Treasury rate at each future
moment. Knowing the probability distribution allows useful quantities such as the expected
Treasury rate at each future moment to be calculated. Each Treasury rate path, in turn,
influences future deposit rates and remaining balances, and the expectations of these
variables can be calculated. A third component of cash flows, deposit acquisition and
servicing costs, net of depositor fees, is also specified in these models.
The way uncertainty is handled in the model also affects the way discounting to calculate
present value is handled. In models with uncertain cash flows, there are two ways to
approach discounting for calculating present value: (1) Expected cash flows are discounted
using a discount rate that includes a risk premium reflecting the riskiness of the cash flows.
For example, when valuing default-free cash flows, the appropriate discount rates are the
rates on Treasury securities maturing at the same date as the cash flows. These rates may
differ by maturity, reflecting a “term,” or maturity, premium, which is a premium for interest-rate
risk.10 When valuing corporate bonds, the discount rate will include a premium for interest rate
risk and a premium for default risk. OAS models for mortgage-backed securities and NMDs
are of this type. (2) Expected cash flows are adjusted to a “certainty equivalent” cash flow by
9
These models include Selvaggio (1996), Hutchison and Pennachi (1996), Jarrow and van Deventer
(1998), and O’Brien (2000).
10
Theories differ as to how term premia may vary with maturity. A summary can be found in Fabozzi
(2000). Using Treasury rates for discounting default-free cash flows, however, implies no stance with
regard to theories of term premia.

subtracting a risk premium, and certainty equivalent cash flows are discounted at a riskless
rate. This is the method used in option-pricing models, such as the Black-Scholes model of
stock option pricing and three of the advanced NMD models.11 The more modern terminology
for option-pricing models is “contingent claims” models.12
Finally, the effective (final) maturity of the cash flows must be specified. In the advanced
models, this could be infinite, since the deposit balance equations may never result in a zero
balance. In practical implementation of the models, however, the horizon is assumed to be
long, but finite, such as 30 years, after which the present value of future cash flows is so small
as to be negligible.
Thus, a fully-specified NMD model consists of the following:
• A clear and rigorous specification for how uncertainty enters the model. Stated
another way, there must be one or more mathematical equations for calculating
the possible future paths of any variable that is a fundamental source of
uncertainty. In the current NMD models, this means a mathematical/statistical
description of possible paths of the short-term Treasury rate.
• A deposit rate equation, which specifies how the deposit rate depends on the
Treasury rate (and other factors such as lagged deposit rates).
• A deposit balance equation, which specifies how the remaining deposit balance
depends on the deposit rate, the Treasury rate and other factors such as
lagged deposit balances.
• An assumption about acquisition and servicing costs, net of any fees paid by
depositors.
• A theoretically sound method of discounting, either discounting expected cash

flows at a risk-adjusted rate or discounting certainty-equivalent cash flows at a
riskless rate.
11
Hutchison and Pennachi (1996), Jarrow and van Deventer (1998), and O’Brien (2000).
12
This certainty-equivalent interpretation of contingent claims models can be found, for example, in
Trigeorgis (1997), particularly p. 103.

• Estimated model parameters (fixed values in equations, such as the coefficient
on the Treasury rate in a deposit balance equation) from a valid database of
credit union deposits.
We have added the last ingredient in the interest of practical implementation of these models.
Fully-specified models without reliably estimated parameters are of little use for estimating
NMD values.
B. Economic Rents as the Source of NMD Premia
Although the basics of valuation theory, as discussed above, are important for an
understanding of various methods of NMD valuation, one can understand these quite well and
still remain confused about the source of NMD value: why do NMDs have premia? For
example, why would a $100 deposit be valued in an NEV context at $95? The language is
already confusing, because in securities markets we would describe the $5 difference as a
discount. A corporate bond, for example, might be issued not at par ($100), but at a discount
($95) if its coupon is less than the required coupon on comparable bonds. This confusing
terminology is a clue that something is different about NMD “premia.”
In fact, the corporate bond analogy is fundamentally flawed. The corporate bond discount
occurs whenever the coupon rate is lower than the required rate on comparable bonds. In
contrast, the NMD premium exists when the deposit pays exactly the required rate on
comparable deposits. For example, a financial institution can issue a $100 share draft paying
the required rate on comparable share drafts and immediately value the deposit at $95 for
NEV purposes. The $5 difference is not a discount at all, but an additional, implicit asset
known as an “economic rent.” Economic rents may be defined as profits that are not expected
to be competed away.
The existence of economic rents is due to the special nature of retail deposits in banking
institutions.13 These deposits are essentially free of credit risk for the depositor due to federal
insurance guarantees, and yet the deposits typically pay rates lower than the Treasury rate,
which is the market rate on a comparable-risk investment. The market value of a financial
instrument that pays less than the rate on a comparable-risk investment will reflect this rate
13
Important articles in this framework are Hutchison and Pennachi (1996), Jarrow and van Deventer
(1998), and O’Brien (2000).

differential, or spread. This spread, less the institution’s costs of acquiring and servicing the
deposit, is often called an economic rent in the literature. It can be defined as
 Interest that could be received  Interest   Non − interest 

Time t Economic Rent =   −   −  
 from investing deposit   Expense   expense 
= (rt − it − c t )Dt −1 (1)
where r is the Treasury rate, i is the interest rate paid to depositors, c the is cost of acquiring
and servicing the deposit expressed as a rate, and Dt-1 is the deposit balance at the end of the
previous period. Since these rents occur in the future, they must be discounted in order to
estimate their present value. The appropriate discount rate is the Treasury rate, if the cash
flows are assumed to have no credit risk.
The deposit liability can then be expressed as a net liability after subtracting the net present
value of the economic rent,
 Net Present Value 

 Face Value   
Deposit Liability Value =   −  of  (2)
 of Deposits   Cash Flows 
 
In this generalized formulation, the net economic value of the deposit is the difference between
the face value and the net present value of the economic rents expected to be earned on that
balance in the future. The present value of the economic rents is referred to in this literature
as the deposit “premium” and often expressed in percentage terms by dividing it by the initial
deposit balance.14 This approach is consistent with the NEV concept.
The deposit premium due to economic rents can also be computed by discounting the deposit
cash flows, without explicitly including the interest on the comparable-risk investment. Deposit
cash flows consist of changes in deposit balances, interest paid to depositors and non-interest
expenses. The cash flow (received by the institution) in time period t can be written
14
For example, the net present value of the economic rent might be 4 percent of the face value of the
liability. For a deposit of $100, the deposit liability value is then $100 - $4 = $96. Note that it is
possible for the deposit premium to be negative if an institution does not keep its deposit rate lower
than the comparable-risk, or Treasury, rate. Some negative deposit premiums have been estimated in
empirical studies, although positive premiums predominate.

 Change in   Interest   Non - Interest 
Period t Cash Flow =   −   −  
 Deposit Balance  Expense   Expense  (3)
= (Dt − Dt −1 ) − (it + ct )Dt −1
The net liability of the deposit can then be written
 Net Present Value 

 Face Value   
Deposit Liability Value =   −  of  (4)
 of Deposits   Cash Flows 
 
In this “deposit cash flow” framework, the “Net Present Value of Cash Flows” is equivalent to
the “Net Present Value of Economic Rents” if the cash flows are discounted at the Treasury
rate; the result is the same as if the deposit had been invested to earn the rate on Treasuries.
In other words, one can attribute the deposit premium to economic rents and then calculate
that premium in either of two ways: with an explicit interest rate spread in the numerator of the
calculation as in (1), or with deposit cash flows (including changes in deposit balances) in the
numerator as in (3). In both cases, the Treasury rate must be the discount rate, because this
rate reflects both the credit risk of the deposit (zero) and the interest rate risk of the deposit
(that of a comparable-maturity Treasury). We take some pains to demonstrate this below,
because a thorough understanding is useful for discussing the choice of discount rate in NMD
valuation.
C. Valuing Deposit Liabilities Based on Economic Rents
We demonstrate the two ways of defining cash flows in the economic rent equation for
calculating the deposit premium. For simplicity, we demonstrate this for the simplest case of
constant interest rates and certain deposit balances for two periods. This two period model
assumes the deposit is withdrawn at the end of the second period. First, we define the
variables:
Vd = deposit liability value.
D0 = par value of deposit.
D1 = deposit remaining at the end of the first period, which is returned to the depositor
at the end of the second period, assumed known with certainty in this derivation.
i = credit union deposit rate, assumed constant in this derivation.

c = credit union servicing and other costs as a percentage of the deposit balance,
assumed constant in this derivation.
r = Zero–coupon Treasury rate, assumed constant over time in this derivation, and, for
discounting purposes, the one-period rate and the two-period rate are assumed
equal.
We now examine the two ways of incorporating the value of the economic rents on deposits
into the calculation of the economic value of the deposit.
A. Compute the net present value of the rent directly by discounting the cash flows due to
the interest rate spread at the Treasury rate. The deposit liability value is the par
amount of the deposit less the net present value of the rent. This is the method
typically found in the research literature.
 D (r − i − c ) D1 (r − i − c ) 
V D = D0 −  0 +
(1 + r )2 
(5)
 1+ r
We show equation (5) in the same format as equation (2).15 The value of the deposit in
equation (5) equals the face value of the deposit (the first term on the right-hand side)
minus the present value of the economic rents for the two periods (the bracketed
second term).
B. Discount the deposit cash flows at the Treasury rate. The result is the present value of
the deposit liability. In this method, the value of the rent is implicit in the calculation
due to discounting the cash flows at a higher rate than the deposit rate.
 ( D − D0 ) − D0 (i + c) (− D1 ) − D1 (i + c) 
V D = D0 −  D0 + 1 +  (6)
 1+ r (1 + r ) 2 
T
 Dt −1 (r − i − c) 
15
The T-period form of equation (5) is: V D = D0 − ∑
t =1  (1 + r ) t
.


We show equation (6) in the same format as equation (4).16 17 The value of the deposit in
equation (6) equals the face value of the deposit (the first term on the right-hand side) minus
the net present value of the cash flows (the bracketed second term). In the second quotient of
the bracketed second term, note there is no D2 because the deposit remaining at the end of
the second period is zero.
To demonstrate that the two methods (equations (5) and (6)) are equivalent, in equation (6) we
first combine the term D0 (within the brackets) with the first quotient:
 D (1 + r ) + (D1 − D0 ) − D0 (i + c ) (− D1 ) − D1 (i + c ) 
V D = D0 −  0 +  (7)
 1+ r (1 + r )2 
Continuing, we simplify and reorganize the first quotient:
 D (r − i − c ) + D1 (− D1 ) − D1 (i + c ) 
V D = D0 −  0 +  (8)
 1+ r (1 + r )2 
We move the term D1 from the first quotient and combine it with the second quotient:
 D (r − i − c) D1 (1 + r ) + (− D1 ) − D1 (i + c) 
V D = D0 −  0 +  (9)
 1+ r (1 + r ) 2 
Finally, we simplify and reorganize the second quotient:
 D (r − i − c ) D1 (r − i − c ) 
V D = D0 −  0 +
(1 + r )2 
(10)
 1+ r
Equation (10) is the economic value of the deposit as described in A in equation (5). Thus, A
and B are equivalent (either can be derived from the other).
16
Note that equation (6) may be expressed alternatively as:
( D0 − D1 ) + D0 (i + c) D1 (1 + i + c)
VD = +
1+ r (1 + r ) 2
T
 ( D − Dt ) + Dt −1 (i + c) 
17
The T-period form of equation (6) is: VD = ∑  t −1 
t =1  (1 + r ) t 

It is useful to reiterate why economic rents exist on NMDs but not on other liabilities such as
corporate bonds, or even corporate equity. The reason is that non-deposit liabilities and equity
do not earn economic rents. NMDs have an arbitrage profit potential those other sources of
funds lack. It is typically assumed that debt liabilities and equity are priced in a competitive
market. If a firm were able to issue a bond at less than the Treasury rate and invest the
proceeds in a Treasury security, the firm would have an arbitrage opportunity (a “money
machine”). In a competitive market, these arbitrage profits would be competed away as other
firms entered this lucrative business – the supply of corporate bonds would increase, with
prices falling and yields rising, until no arbitrage profits were available. A deposit, however, is
a “money machine” paying rates lower than Treasuries without causing a competitive market
response that eliminates the interest rate spread. Whether a particular financial institution
actually reinvests deposits in Treasuries is not relevant to the valuation argument. The
question is: What must be the market value of a liability that always pays a rate less than an
equivalent-risk investment opportunity? The arbitrage profit is an inherent part of the value of
that liability. 18
Could the economic rent be calculated by assuming the deposit were reinvested in a different
asset than a Treasury security? How about a mortgage-backed security? How about a new
addition to the building? Such investments would not provide economic rents, that is, they
would not be arbitrage trades, because the investments are riskier than the deposit. It is not
arbitrage to issue a riskless deposit and invest in a risky investment. The valuation of deposits
by computing the arbitrage profit, or economic rent, is fundamental.
A final point: It is not clear that all of the methods used by credit unions to “value deposits” are
really intended to be economic value calculations. Discounting deposit balances only, for
example, is not. To calculate economic value, deposit interest (and servicing and other costs)
must be included in the cash flows.
18
As noted by Jarrow and van Deventer (1998): “In economic terms, the net present value is the
maximum premium above the dollar amount of deposits that a rational bank would bid to purchase the
demand deposit franchise from another bank. It is analogous to the net present value computation
one would make for an investment project in traditional capital budgeting problems. It represents a
rent for the privilege of issuing demand deposits. Note that this value is determined independently of
the use to which these funds are put within a bank.” (p. 256)

D. Choosing the Discount Rate
With this understanding of the theory of the economic value of deposits, we are in a better
position to discuss the appropriate discount rate for credit unions to use. It is clear that if a
credit union intends to incorporate the value of economic rents in the value of deposit
liabilities, then the discount rate must be the rate on a comparable-risk investment. If the NMD
is assumed to be default-free, then a Treasury rate is the appropriate discount rate.
Would some other rate would be appropriate? How about the deposit rate or the rate on
comparable deposits? Discounting the deposit cash flows at a deposit rate is appropriate if
the goal is to calculate the value of the liability without rents. Think of an analogy to a non-
financial corporation that has issued a bond for $100 with a coupon of 5%. At issuance, the
bond is worth par, and we can calculate this present value by discounting the bond cash flows
at 5%. This does not provide any idea of what the firm may earn by investing the proceeds of
the bond issue. Later if the yield of the firm’s bonds increased to 6%, we could compute the
market value of the same bond by discounting the cash flows at 6%, and the bond would be
worth less than $100. This would be the value of the bond, alone, not the value of the bond
less economic rents. Similarly, if we discount the deposit cash flows at a deposit rate, which,
by the way, should be the rate that the credit union has to pay today (not when the deposit
was first accepted), we have a valid “mark to market” of the liability, but we have neglected to
include the extra component of deposit valuation, the rents. If we want rents, and thus we
want to define the economic value as conventionally defined, we have to use the framework
represented by equation (5) or (6).
Is an alternative cost of funds an appropriate discount rate? The discount rate should be the
rate on an investment with the same risk to a depositor as the deposit. To the extent that
credit union retail deposits have no credit risk for the depositor, due to the insurance
guarantee, the appropriate comparable-risk investment is a Treasury security and the
appropriate discount rate is the Treasury rate. Other rates, such as the London Interbank
Offered Rate LIBOR, rates on secondary market certificates of deposit (CDs) and the Federal
Home Loan Bank (FHLB) regional cost of funds, are appropriate only if it can be argued that
credit union NMDs have approximately the same credit risk as Eurodollar deposits, uninsured
CDs and loans to thrift institutions, respectively.19 Although these rates may be appropriate
19
Mays (1997) notes that these alternative costs of funds are often used by banks and thrifts in NMD
valuation.

measures of alternative costs of funds for some institutions, the discount rate in the
economic rent calculation is not a cost of funds. The purpose of the exercise is not to
compare the NMD deposit rate with the rate on a jumbo CD in order to show how much
“cheaper” it is to raise funds with deposits. For the purpose of valuation for the NEV
calculation, that is an “apples and oranges” comparison; the two financial instruments have
different credit risk. Because the alternative costs of funds are typically higher than Treasury
rates (as they must be on instruments of greater credit risk), the value of economic rents on
default-free NMDs will be understated.
Option-adjusted spread models can be based on virtually any benchmark rate, although a
short-term Treasury rate is commonly used as in Selvaggio (1996). A spread is added to the
benchmark rate that forces the model price to be the same as the market price of the security
or deposit in question. As discussed elsewhere, we do not recommend that credit unions use
these models because of the absence of a market for credit union deposits, and the resulting
reliance on untested (in the credit union market) proxy measures of the option-adjusted
spread.
It is recommended that zero-coupon Treasury rates (rates which are specific to the date on
which a cash flow is anticipated) be used to discount estimated cash flows at the dates in the
NMD model. Zero-coupon Treasury rates have the same default risk as the deposit and also
have any premium for interest rate risk that may be specific to longer maturities. Treasury
zero-coupon rates are readily available to most credit unions. If zero-coupon rates are not
available, an acceptable proxy is the Treasury note or bond yield to maturity for the maturity of
each cash flow in the model.
Should servicing and other costs be included in the discount rate? No. The correct approach
is to include such costs as cash outflows in the numerator of the present value calculation.
E. Assumptions Underlying Specific Methods
We now turn to the specific methods available from vendors and published in the research
literature. The Appendix contains a more detailed and technical description of the different
methods and the underlying formulas (where available), while the following sections discuss
and contrast their most important features.
Table 2 summarizes the different assumptions underpinning the methods in the context of a
fully specified model as discussed above.

Table 2. Comparison of Assumptions Underlying Different Valuation Methods for NEV
————————————————————————————————————————————
Description: This table indicates how various models satisfy the requirements of a fully-specified
model..
————————————————————————————————————————————
Deposit Remaining Includes
Market Rate Tied Balance tied Acq. &
Interest Rate to Market to Market Svc. Effective
Method Uncertainty Rate Rate Costs Discounting Maturity
CU/ALM None No No No None 1 Month
User-
1
IPS-Sendero None No No No None Defined
2 2
PROFITstar Scenarios No No No None Long
CUNA Model User-
3 4
Management Scenarios No Yes No Yes Defined
C. Myers Scenarios No No No None Zero
Hutchison & Vasicek Treasury
5
Pennacchi Model Yes Yes Yes rate 30 years
Jarrow & van
Deventer Heath-Jarrow-
(Kamakura Morton Treasury
5
Corporation) (HJM) Yes Yes Yes rate 30 years
McGuire
Performance Scenarios or Alternative
Solutions Random Yes Yes Yes cost of funds 17+ years
Cox-Ingersoll-
Ross (CIR) Treasury
5
O’Brien Model Yes Yes Yes rate 30 years
OTS LIBOR +
6
(2001) None Yes Yes Yes OAS 30 years
Eurodollar +
5
Selvaggio CIR No Yes Yes OAS 30 years
Source: Original papers and NCUA internal documents.

1
Note: User can override and supply NMD value.
2
User can supply balance decay rates. Discounting at offering rate.
3
One user choice is based on an OTS table developed by Forin and Associates.
4
User-defined or yield-curve based discount rate
5
The Vasicek model generates random, mean reverting spot (zero-coupon) rates. The CIR model is
similar but does not allow negative rates. The HJM model is based on random evolution of the forward rate
curve, which then determines the random evolution of spot rates.
6
One future path of spot rates is assumed based on current forward rates.
Several of the vendor methods ignore how interest rates may change and fail to fully specify
NMD cash flows. Models that do take into account future changes in rates allow a set of
scenarios, assume a specific random interest rate model, or use the forward rates implied in
the current yield curve. The option-adjusted spread (OAS) approach incorporates the effect of

the options embedded in the deposits by adding a risk premium to the discount rate; it also
requires that the cash flows being discounted reflect how changes in interest rates affect those
cash flows.
The key credit union behavioral assumption is the relationship between market interest rates
and deposit rates. In those methods where it is modeled explicitly, most assume that the
change in the deposit rate is the same regardless of whether interest rates are increasing or
decreasing. Since all empirical evidence has shown that credit unions and deposit-taking
institutions in general tend to react differently to increases in market rates than decreases in
market rates, this may have implications for the accuracy of the methods. 20
It would appear that the option-related features of NMDs are less prevalent in credit unions
than in other deposit-taking institutions. According to Tripp et al. (1997), credit union deposit
rates and balances are less interest rate sensitive than those of banks and thrifts. To the
extent that this is the case, the cash flows associated with credit union NMDs would display
fewer option-like characteristics. By this, we mean that the cash flows and deposit balances
would vary less with the different interest rate scenarios. This has important implications for
whether the more complex valuation methods are cost-effective for credit unions.
The key depositor behavioral assumption regards how deposit balances evolve through time.
This is handled by assuming an effective maturity for the deposits, by estimating a retention
(or decay) rate of the remaining balance (so that the same proportion each month is assumed
to be withdrawn), or by estimating a relationship between deposit balances, deposit rates,
market rates, and other variables.
In the interests of better understanding the nature of the different methods, they have been
loosely grouped into three different types. It is important to note that most are not uniquely
characterized; thus, for example, the contingent claims methods all contain elements of the
present value method. The various methods and approaches to valuing NMDs are discussed
in the following sections, in terms of how each one addresses the two main problems outlined
above, how to deal with the “maturity” and uncertain cash flows of NMDs. The cash flow
problem has really two parts: how does the interest rate on deposits vary with market interest
20
See, for example, Tripp, Lacewell, and Kenny (1997) for credit unions and O’Brien (2000) for banks
and thrifts.

rates, and how do deposit balances change with deposit rates, market rates, and other
economic or institution-specific variables.
F. Simple Vendor Methods
There are two alternative simple approaches to dealing with the maturity of non-maturity
deposits that are used by several commercial vendors as part of a more comprehensive asset-
liability management system. One is to assume that all non-maturity deposits are extremely
short-term in maturity (one month or less); the other is to assume that they are all very long-
term in maturity. The two assumptions have very different implications for calculating NEV
and interest rate risk.
The assumption that non-maturity deposits are extremely short-term in maturity is the
equivalent of saying that deposits are fully demandable and will be withdrawn every month.
Another way of stating this is that NMDs do not have any economic rents, so that the second
part of equation (2) is zero.21 This will tend to overstate an institution’s interest rate risk and
understate its NEV in a period of rising interest rates, while the reverse will be true if rates are
falling. On the other hand, assuming a very long maturity is the equivalent of assuming a very
high retention rate for deposits. Consequently, there is the potential for a large economic rent
component to the deposit premium. It also implies that the deposit balances are not sensitive
to changes in the level of deposit rates or the spread between deposit rates and market rates.
The range of assumed effective (final) maturities used by many of these vendor methods
raises the issue of what is the appropriate choice. The choice of effective maturity cannot be
made without first specifying the model that is being used, and in particular how NMD cash
flows are modeled. The approaches to dealing with future cash flows from the deposits vary
widely. Some of these models do not attempt to specify future deposit balances nor deposit
interest and servicing cash flows, and consequently they do not calculate a present value
based on explicit cash flow assumptions. Such methods are unlikely to provide a reliable
estimate of deposit premiums and durations. The failure to specify future cash flows and
calculate present values for deposits is inconsistent with the treatment of other assets and
liabilities in NEV calculations. We address the choice of effective maturity in such methods in
Chapter IV.
21
Hutchison and Pennacchi (1996) show this and discuss the implications for deposit valuation.

G. Present Value Methods
All the different present value methods apply traditional discounted cash flow analysis in order
to estimate the value of non-maturity deposits. However, they differ in their approach to
dealing with the maturity and the option-related features of NMD cash flows.
1. OTS Net Portfolio Value Model
The approach taken by the Office of Thrift Supervision (OTS) towards non-maturity deposits in
its Net Portfolio Value Model (hereafter OTS NPVM) is quite straightforward. It allows for
changing cash flows through dependence on market and deposit rates, but since only one
interest rate scenario is considered, it effectively is a static discounted cash flow approach. It
calculates the static discounted cash flow for each of the four types of account, using monthly
projected cash outflows. Zero-coupon LIBOR rates plus an OAS are used for discounting over
a 30-year (360 month) horizon.22 NMDs are assumed to have a potentially long maturity, but
the retention rate for aggregate deposit balances that determines the duration is obtained
empirically either from the specific institution being valued, or from industry-wide estimates. If
the institution’s own retention rate is used, the difference between that number and the
industry-wide retention rate is also incorporated, and the difference between the two is
assumed eventually to converge to zero.23
The monthly cash flows result from the net effect of interest payments on the balances,
withdrawals, and non-interest expenses associated with the accounts. The rate paid on
deposits is modeled as a function of recent deposit rate levels and recent market rate
changes, and whether the deposit rate is above or below the long-term “equilibrium” rate on
such deposits. Changes in deposit balances are similarly modeled as a function of the spread
between deposit rates and market rates, the sensitivity of that institution’s deposit rate to
changes in market rates, the institution’s retention rate and the industry-wide retention rate.
22
The OTS discount rate is LIBOR plus a spread that “calibrates demand deposit prices to observed
prices for deposit purchase/assumption transactions.”
23
In April 2001, the OTS announced changes to the specific equations that comprise their model, in
terms of both the functional form and the value of some of the parameters. The underlying philosophy
and objective of the model remain the same, however, and do not alter any of the conclusions drawn
in the present document. The two versions are both presented and discussed in the Appendix.

The method is a “top-down” method, in the sense that most parameter values (e.g., non-
interest cost rates, the sensitivity of deposit rate changes to changes in market rates, etc.) are
estimated from data from all thrifts and applied to all thrifts. Some variables do vary across
institutions, such as the starting level of account balances and the rate paid on a particular
type of account. An important feature of this model is that it is effectively imposed on all thrift
institutions. The OTS estimates many of the parameters, and these are used at all institutions
regardless of size or other distinguishing characteristics. This has important implications, both
positive and negative. By estimating parameters that can be used by all thrift institutions, the
OTS retains control over the process and ensures that the necessary statistical estimation
procedures are conducted in the proper manner. It also means that appropriate estimation
procedures are effectively placed within the reach of even the smallest and least sophisticated
institution. However, it also means that the OTS is effectively mandating a “one size fits all”
approach, with little or no consideration of the characteristics that distinguish the regional
deposit markets, or that differ from small to large thrifts.
One important implication of adopting the OTS methodology for the valuation of credit union
NMDs is that NCUA would be responsible for estimating all of the industry-wide parameters
that are used, for keeping the parameters up-to-date, and for communicating any changes in
the model and its inputs to all insured credit unions. A legitimate question to ask, therefore, is
whether NCUA wants to enter the business of estimating model parameters and monitoring
how they are used by credit unions. This is discussed in more detail in Chapter V.
2. McGuire Performance Systems
The approach of McGuire Performance Solutions, on the other hand, uses data from a single
bank, thrift, or credit union, and estimates statistically the rate at which deposit balances
decline for different interest rate scenarios. Thus in a sense it can be said to be at the other
end of the spectrum from the OTS approach, since only data from a single institution is
incorporated. All parameters are specific to the institution whose deposits are being valued,
and are obtained from long time-series of observations of both aggregate deposit balances
and a sample of individual accounts. The precise functional form and parameter values are
unique to each institution, and depend on the availability of extensive historical data on all
types of deposits, costs, and other factors. The underlying philosophy of this method of
valuing the deposit premium associated with NMDs is to model the core deposit behavior at
the specific institution using a set of simultaneous equations for deposit rates and balances.
The variables used in the equations (as well as the functional form of the equations) are

selected using a statistical discovery process, but generally start off with a list of candidates
that include lagged deposit balances, rates paid on deposits, spreads between deposit rates
and market rates, trend variables, seasonal factors, and general and local economic data.
The chief advantage of this approach is that each valuation is uniquely defined for a given
institution. Unlike the OTS NPVM, the McGuire approach does not confine itself to a single
possible interest rate path, but instead looks at many different interest rate scenarios with
different cash flows for each. In this way, the options embedded in NMDs are explicitly taken
into account.
A disadvantage of the McGuire method, in the context of credit unions, is the requirement for
long time series of data for many individual accounts of each account type. Probably few
credit unions would have such data available. In addition, this method allows a limited number
of interest rate scenarios to be utilized in the valuation process.
3. Selvaggio, OTS and OAS
The approach adopted by Selvaggio adjusts the basic present value method for what is known
as the option-adjusted spread (OAS). This is the spread that must be added to the zero-
coupon Treasury in order that the observed market value of core deposits equal the value
produced by the model. Empirically, he finds that the OAS needed to equate the present
value of economic benefits with observed premia in the banking market for NMDs is about
1,100 basis points or 11%. The approach to the maturity question is similar to that of OTS, in
that a 360-month horizon is used. Static discounted cash flow is not used, since 300 interest
rate paths are simulated; consequently both the cash flows and the discount rate arrived are
adjusted to reflect the impact of the embedded options. Finally, the expected change in
deposits is modeled as a function of recent deposit levels, current market rates, a time trend,
and seasonal dummies.
In its most recent version of the NPVM, the OTS adds an OAS to the zero-coupon LIBOR rate.
The spread reported in April 2001 is 12 basis points.
This approach to valuing NMDs has some problems when applied to credit union NMDs.
Since the goal of the present exercise is the reverse of what Selvaggio and OTS do, namely to
value the deposit premia rather than infer the appropriate discount rate, to implement this
process would require knowledge of the appropriate OAS on the valuation date. Since there is
not an active market in the deposits of credit unions, this would require making some key
assumptions. This point is considered in more detail, in Chapter V.

H. Contingent Claims Methods
In the last few years, three methods have been proposed that use a contingent claims
approach to the problem of valuing non-maturity deposits. These are the models of Hutchison
and Pennachi (1996), Jarrow and van Deventer (1998), and O’Brien (2000).24
A key difference between the contingent claims valuation methods and the present value
methods of the previous section is how future interest rates are handled. Present value
methods that use static cash flows, such as the OTS NPVM, incorporate only one interest rate
path, that implied by the forward rates embedded in the current yield curve; this may be
loosely described as an “expected” path for interest rates.25 While the cash flows estimated
for this expected interest rate path may reflect the options embedded in the NMDs, they come
from only one possible path out of many. Contingent claims valuation methods, on the other
hand, involve consideration of hundreds or thousands of possible interest rate paths. This is
because the option-related features of NMDs mean that future cash flows are not known with
certainty, but depend on the particular path that interest rates (both market rates and deposit
rates) take. By running simulations of many possible interest rate paths, the cash flows
associated with the NMDs can be determined in each path and an average over all paths
determined. 27 Thus, the resulting valuation reflects the impact of the embedded options, and
the likelihood that the options will be “exercised” over many different interest-rate paths. This
gives a more accurate estimate of the impact of the embedded options on the cash flows and
value of the NMDs.
24
Jarrow and van Deventer (2000) and O’Brien (2000) can be categorized as arbitrage-free models,
while Hutchison and Pennacchi (1996) can be categorized as an equilibrium model. Hutchinson and
Pennachi determine the deposit rate endogenously (within the model). The others assume a deposit
rate equation without derivation.
25
This description is more for convenience than precision; there is broad consensus in the economics
literature that forward rates generally differ from expected future spot rates due to positive or negative
term premia in default-free bond prices reflecting interest-rate risk.
26
For example, the valuation results from the O’Brien model in Chapter VI, reflect 1,000 simulated
interest rate paths. Simulation is not required in all cases. The Hutchison and Pennachi model and
one version of the Jarrow and van Deventer model have closed-form solutions for the probability
distribution of the NMD cash flows.
27
For example, the valuation results from the O’Brien model reflect 1,000 simulated interest rate paths.
Simulation is not required in all cases. The Hutchison and Pennacchi model and one version of the
Jarrow and van Deventer model have closed-form solutions for the probability distributions of future
NMD cash flows.

All three methods calculate the present value of the economic rents over very long horizons,
similar to OTS and Selvaggio. They all model the future path of the rate paid on deposits as a
function of future market rates. However, there are some important differences. Both
Hutchison and Pennacchi (1996) and Jarrow and van Deventer (1998) assume that the
change in deposit rate due to a change in market rates will be the same whether rates are
rising or falling. O’Brien, on the other hand, explicitly incorporates the well-documented
phenomenon that the spread between market rates and deposit rates tends to widen in
periods of rising rates and narrow in periods of falling rates. O’Brien is also the only one of the
three methods that incorporates local market-specific variables in the estimation of future
changes in deposit levels: as well as being affected by the deposit rates, market rates, and
recent deposit balances, the balance equation includes nominal income for the region in which
the bank does business. This is an important feature of community banks and credit unions
that has long been ignored by most attempts to model NMDs, namely how the institution is
affected by the regional as well as the national economy.
I. Conclusions
The simple vendor methods that do not incorporate all NMD cash flows are not designed to
estimate NMD value in accordance with standard theory. At best, they can be used by credit
unions as a cost-beneficial alternative to more complex models. In this case, the effective
maturities must be carefully chosen, as discussed in Chapter IV.
Many of the more recently developed methods, particularly the contingent claims methods, are
quite complex. To adopt them requires an understanding of interest-rate contingent claims
theory, how it has been adapted for the valuation of NMDs, and how to use a specific method
to value the NMDs of a specific credit union. There is a danger of incorrectly applying a
method, of incorrectly estimating the required input parameters, and of a false sense of
security that the use of a complex model can sometimes bring. The old computer
programming saying “Garbage In, Garbage Out” (GIGO) is just as applicable in the use of
valuation models. Complex, sophisticated valuation methods should be used only by those
who have the necessary expertise and understanding to be aware of the pitfalls that may arise
in their use.
A key difference between some of the methods is whether they include growth in future
deposits. For example, the OTS NPVM does not include new deposit balances, but models
the gradual decline of the initial deposit. O’Brien, on the other hand, models future total

balances, namely the remaining portion of current balance plus expected growth in deposits.
This can lead to material differences in valuation of the same deposit. Both are valid, in the
sense of being consistent with the assumptions of the particular model. If credit union
deposits are being valued for the purposes of NEV, it is appropriate not to include expected
growth in balances, since the aim of NEV is to estimate the economic value of the existing
balance sheet without the effects of new business.
The present value methods do not compute the value of economic rents in accordance with
the theory in the research literature unless the discount rate is chosen to reflect the credit risk
of the deposit. The use of other rates, such as rates representing alternative costs of funds,
are not recommended in NEV applications unless the user is making the explicit assumption
that NMDs have the same credit risk for depositors as the alternative source of funds.

III. CREDIT UNIONS VS. BANKS
Further complicating the issue of valuing NMDs, as it relates to the present study, is that the
properties and behavior of non-maturity deposits at credit unions are held by some industry
observers to be substantially different from those of banks and thrifts (hereafter banking
institutions). Most of the methods discussed in Chapter II were originally designed with more
traditional banking institutions in mind, and it is important, therefore, to determine whether the
unique characteristics of credit unions that distinguish them from banks render the use of any
these methods inappropriate.
The next section discusses the similarities and dissimilarities of credit unions and banking
institutions. The implications of those dissimilarities for the present problem of applying
existing methods to value the NMDs of credit unions is considered in section B.
A. Similarities and Dissimilarities
1. Structural Differences
In structural terms, credit unions and banking institutions overlap and are broadly similar in
many areas. Both take in consumer deposits and make consumer loans. Both are regulated
by state and federal agencies. Both are exposed to credit risk and interest rate risk. However,
the study of credit unions by the U.S. Department of the Treasury in 1997 noted that there are
five main characteristics that distinguish credit unions from banks and thrifts:28
1. Credit unions are member owned, member-directed depository institutions.
2. Credit unions rely on unpaid, volunteer boards of directors elected by, and drawn
from, each credit union’s membership.
3. Credit unions are not-for-profit organizations.
4. Credit unions have a public purpose, “to make more available to people of small
means credit for provident purposes.”
28
U.S. Treasury (1997), p. 1.

5. Credit unions face limitations on their membership, usually referred to as a
“common bond” requirement.
Perhaps the most important of these distinctions, as related to the present study, are the first
and third. The first distinction means that credit unions do not issue stock to raise capital,
instead they rely on accumulated retained earnings for their net worth. This has important
implications for regulators, since a troubled credit union whose capital is in danger of dropping
below the level considered adequate cannot simply raise additional equity capital by selling
more shares. Nor do credit unions have access to the debt markets or markets for large
commercial deposits.
2. Behavioral Differences
The structural differences outlined above mean that credit unions do not necessarily respond
in the same way as banks or thrifts to market or economic events. For example, characteristic
number three means that credit unions tend to price their loan and deposit products differently
than banks and thrifts. Whereas a bank or thrift will seek to maximize its profit for
shareholders by charging more for loans and paying less on deposits, a credit union will seek
to return part of its profits to its members by charging them less on loans and paying more on
deposits. This is born out in Table 3, where we see for example that in January 2001 new
auto loans were 140 basis points cheaper than comparable loans at banks and S&Ls, and
share draft accounts paid 86 basis points more than their bank equivalents.
Table 3. Comparison of Rates at Credit Unions and Banks

Description: This table compares the level of deposit and loan rates at credit unions with those of similar
products at banks and thrifts
January 2001 Credit Unions Banks, S&Ls
Loan Rates
New Auto 8.12% 9.52%
Personal 13.32% 15.23%
Credit Card 12.63% 15.46%
Savings Rates
Share Draft 1.66% 0.80%
Money Market Account 4.21% 2.06%
One-year CD 5.92% 4.78%
Source: CUNA website (www.cuna.org), April 23, 2001.

The spread between credit union and bank rates is not constant, however; Tripp et al. report
that the difference between bank and credit union savings rates narrowed considerably in the
mid-1990s as credit unions were slower to raise their rates in response to rising market rates.
The relationship between credit union rates and other rates can be seen in Figure 1, which
compares the average rates paid on credit union deposit accounts with the yield on 3-month
Treasury bills and 1-month bank CDs.
It is readily apparent that the rate on credit union nonmaturity deposits reacts much more
slowly than bank CDs to changes in the level of interest rates. This can be seen more clearly,
when we examine the correlation coefficients between changes in credit union rates and
Treasury bill rates. Changes in the rate on share drafts have a correlation with changes in the
3-month Treasury bill rate of only 0.19; for regular shares, the correlation is 0.52, and for
money market shares, it is 0.76. For bank CDs, the same measure of correlation is 0.91.
Description: Rates offered in credit union share drafts, regular shares, money market shares, and 1-
year certificates compared to the rate on bank CDs and Treasury bills.
————————————————————————————————————————————
6
Yield (%)
0
Ap 1
Au 2
D 92
Ap 2
Au 3
D 93
Ap 3
Au 4
D 94
Ap 4
Au 5
D 95
Ap 5
Au 96
D 96
Ap 6
Au 7
D 97
Ap 7
Au 8
D 98
Ap 8
Au 9
D 99
Ap 9
0
-9
-9
-9
-9
-9
-9
-9
-9
-9
r-9
r-9
r-9
r-9
r-9
r-9
r-9
r-0
g-
g-
g-
g-
r-
g-
g-
g-
g-
ec
ec
ec
ec
ec
ec
ec
ec
ec
D
3M TBILL 1 Mo. CD Reg. Shares MMShares 1 YR Certificates Share Drafts
Source: CUNA and Board of Governors of the Federal Reserve.
Figure 1. Comparative Yields
The other chief distinction between credit unions and banking institutions is size. While the
number of credit unions is more than the number of banks or thrifts, the average credit union
has only $38 million in total assets; for banks and thrifts the numbers are $668 million and

$700 million respectively.29 Thus, credit unions are at a considerable disadvantage with
respect to banking institutions when it comes to the economies of scope and scale, which
have been shown to be of great importance for deposit taking institutions. They are also at a
disadvantage when it comes to being able to diversify their loan portfolio: because credit
unions can only lend to their members, this effectively places restrictions on their ability to
make loans to a geographically or (for single-purpose credit unions) economically diversified
group of borrowers.30
There are also differences between credit unions and banks in terms of the behavior of the
various NMDs. For example, Figure 2 and Figure 3 show the difference in account balances
and transactions volume of non-maturity deposits at banks and credit unions. Note that for
smaller credit unions, accounts with less than $5,000 represent almost 90% of balances and
over 45% of transactions, whereas for small banks the same account size represents 60% of
account balances and barely 10% of transactions. For both large and small credit unions,
accounts between $1,000 and $5,000 show the most activity (roughly 35% of transactions in
both cases) whereas the same account size for banks sees less than 10% of activity. Theses
differences may have important implications for the pricing and costs of NMD-products at the
two types of institution.
29
Source: CUNA. Data are as of December 1999.
30
See Evans and Shull (1998) for a more detailed discussion of the impact of restrictions on Credit
Unions on their ability to diversify.

————————————————————————————————————————————
Description: This figure shows the distribution of average account balances by size of bank and credit
union.
————————————————————————————————————————————
50%
40%
30%
20%
10%
0%
Under $200 $200 - $499 $500 - $999 $1,000 - $5,000 - $10,000 - $25,000 - Over
$4,999 $9,999 $24,999 $99,999 $100,000
Banks < $200M Assets Credit Unions < $200M Assets Banks > $200M Assets Credit Unions > $200M Assets
Source: Federal Reserve Bank of San Francisco.
Figure 2. Distribution of Account Balances at Banks and Credit Unions
————————————————————————————————————————————
Description: This figure shows the distribution of transaction frequency, depending on account size and
size of bank or credit union.
————————————————————————————————————————————
50%
40%
30%
20%
10%
0%
Under $200 $200 - $499 $500 - $999 $1,000 - $5,000 - $10,000 - $25,000 - Over
$4,999 $9,999 $24,999 $99,999 $100,000
Banks < $200M Assets Credit Unions < $200M Assets Banks > $200M Assets Credit Unions > $200M Assets
Source: Federal Reserve Bank of San Francisco.
Figure 3. Distribution of Transactions at Banks and Credit Unions

There is little hard evidence on whether the NMDs of credit unions display different behavioral
and empirical characteristics than those of banks and thrifts. According to Tripp et al., the
rates on credit union NMD products react to changes in market rates more slowly than those
of banks, and the rate at which deposit balances “run off” or decay is also apparently
different.31 However, the precise degree of differentiation has not yet been determined, as
data on credit union deposit balances is not available in the same amount or quality as is the
case for banks and thrifts. This argues for a study of credit union deposit behavior, and we
return to this point in Chapter VII.
B. Implications for Valuing Non-Maturity Deposits
To the extent that credit unions are deposit-taking institutions that provide NMD accounts and
services, then the application of the various methods of pricing NMDs to those balances at
credit unions should be relatively straightforward. However, to the extent that credit unions are
different, then the issue may not be so easy.
That credit unions are different from banks is not at issue; the Treasury study of credit unions
in 1997 gave the five main differences listed above. The primary concern is whether they are
so different that existing methods for valuing NMDs cannot be modified or applied at all. This
comes down to whether the fundamental differences between credit unions and other banking
institutions violate any of the assumptions underlying the methods. Table 2 on page 21
summarizes these.
While some of the assumptions are clearly motivated by the behavior of banks as opposed to
credit unions (profit-maximization, for example), none of them seem to be so restrictive as to
render their associated method inapplicable for credit unions. The more relevant
consideration is that credit unions are different from banks to the extent that they react to
changes in the economic environment in different ways than banks. This means that the
parameters of the different methods must be adjusted to make them suitable for use by credit
unions. For example, Tripp et al. (1997) document that credit unions were slower to increase
the rates on shares than were banks in the face of rising interest rates; this implies that the
31
An existing deposit tends to decrease as time passes and the depositor withdraws money for
consumption and for alternative savings and investment opportunities. The annual rate of withdrawal
is known as the decay rate. The annual retention rate is the mirror image of the decay rate; it is the
ratio of the remaining balance at the end of the year to the balance at the beginning of the year.

parameter relating changing market rates to changes in the deposit rate would be smaller for
credit unions.
It would appear that the options embedded in NMDs, namely the institution’s option to change
the rate and the customer’s option to withdraw funds, play a smaller role in credit union NMDs.
If that is the case, the component of NMD value attributable to the value of the options will be
smaller, and the use of methods that incorporate these options in balance equations and
deposit rate equations might not make a material difference in NMD valuation for credit unions.
Thus, the main issue regarding the applicability of these methods to credit unions is related to
empirical issues such as the relative rate sensitivity of credit union NMDs to that of banks and
thrifts. These issues will be discussed in more detail in Chapters V and VI.

IV. THE EFFECTIVE MATURITY OF NON-MATURITY DEPOSITS
While they have no stated maturity, deposit balances do have what may be called an effective
maturity, which we define as the assumed final maturity of a deposit. Since deposit balances
may be around for decades, an assumed final maturity is a simplification in the interest of
practical implementation. For example, by estimating and discounting cash flows out to a
horizon of 30 years the OTS NPVM methodology assumes an effective maturity of 30 years.
O’Brien and others make the same assumption in contingent claims models. McGuire
Performance Solutions assumes that the effective maturity of deposits is about 17 years,
based on their empirical estimates of deposit average life. Long maturities are appropriate in
these models because they contain well-specified and empirically-estimated deposit balance
equations. Deposit balances are determined in these models by a retention rate, the spread
between the deposit rate and a market interest rate (typically the Treasury rate), and in some
cases national or regional aggregate income. A deposit balance may never become zero,
although it may become economically insignificant at some point. The only final maturity
requirement is that it be long enough that any economically significant remaining balances
(and the associated interest and servicing cash flows) are included in the present value
calculation. Even with long maturities, these models do not estimate extremely long durations
for NMDs, in part because NMDs are similar to floating-rate instruments.
We make no recommendations as to what the effective maturity should be when relatively

fully-specified models such as these are used by credit unions. We do strongly recommend
that the parameters of such models be estimated using credit union data, since the behavior of
credit union rates and deposit balances may differ substantially from the banking and thrift
deposits for which these models have been used in the past.
The effective maturities that we were asked to recommend in this report apply only to simple
methods that credit unions may use in lieu of more complex and expensive methods. By
simple methods, we mean those that do not attempt to specify the actual cash flows of NMDs.
The cash flows not specified include deposit withdrawals, deposit interest and deposit
acquisition and servicing costs. For example, some simple vendor methods may assume a
final maturity in order to assign NMDs to maturity “buckets” for interest rate risk analysis. In
other cases, the NMD may be treated as a “zero-coupon deposit,” with a constant balance out
to a final maturity, and no specification of interim cash flows. We have been asked to
recommend a reasonable effective maturity in such methods.

A. Duration and Effective Maturity
Simple methods can be satisfactory (cost effective) for small credit unions. For such methods,
the effective maturity is essentially an assignment of interest rate risk to the deposit.
Consequently, the effective maturity, defined as the assumed final maturity of the deposit,
should be based on estimated durations of NMDs found in the literature.
Duration is a measure of interest rate risk widely used in banking and in the analysis of fixed-
income securities. The duration of a security is typically defined as the percentage change in
value for a small change in yield to maturity. Other versions of duration are based on
particular descriptions of how the yield curve may shift. For fixed-income securities with
embedded options, the duration measure should include the effects of changes in cash flows
due to the exercise of embedded options (such as a bond being called) when interest rates
change. A duration measure that includes the effect of changing cash flows is often labeled
“effective duration.”32 The durations reported in much of the recent literature on NMDs are
effective durations.
The choice of effective maturity in simple models of NMDs is essentially an assumption about
interest rate risk. For example, a credit union that assigns one type of deposit to a 3-year
bucket and another type of deposit to an 8-year bucket is assigning greater interest rate risk to
the latter.33 Duration is an appropriate guide for such choices. Our methodology for choosing
the effective maturity in simple methods is to use the effective maturity that produces durations
consistent with NMD durations estimated in the literature and by the OTS. These are
reviewed in the next section.
B. NMD Premia and Duration Estimates in Recent Research
We summarize here evidence contained in recent research or published by the OTS. This
includes measures of premia and duration taken from studies of banks, and the OTS data on
thrift deposits, over different time-periods and using different methods. No such studies of
32
See, for example, Fabozzi (2000). Effective duration is estimated from a pricing model. The
benchmark interest rate is moved up and down a given amount, the cash flows are changed if
warranted (for example, if mortgage prepayments are assumed to change at the new interest rate),
and the new price is computed for the up and down rate movements. Effective duration is the
percentage change in price given the change in the interest rate.
33
To put this in some perspective, assigning an 8-year duration to a NMD is approximately equivalent to
saying that it has the same degree of interest rate risk as a long-term Treasury bond.

credit union deposits exist. Although our main purpose is to estimate effective maturities, we
consider the estimates of premia as well as a rough guide to the reliability of various studies.
Table 4. Premia Estimates for Non-Maturity Deposits

————————————————————————————————————————————
Description: Summarizes published estimates of NMD deposit premia. Reported numbers are medians
from the studies unless otherwise indicated. The premia are expressed as percent of face value.
————————————————————————————————————————————
Period of Non-
Transaction
Author(s) Institution Model MMDA Passbook Interest-
Accounts
Estimation Bearing
Berkovic &
Liang (1991) Failed Banks 1987 – 90 4.9 4.0
Hutchison &
Pennacchi Over 200
(1996) Banks 1986 – 90 6.6 7.0
Janosi et al
(1999) One Bank 1988 – 96 9.7 7.6 171.8
Janosi et al Aggregate
(1999) FED Data 1988 – 95 2.7
McGuire &
Newell
(1997) One Bank 1988 – 95 12.8 0.8 10.2
O’Brien 75 – 100
(2000) Banks 1983 – 94 15.3 10.9
OTS Thrifts 1998 – 01 7.0 2.0 4.0 10.0
Selvaggio
(1996) One Bank 1991 – 95 21.1
Note: The account names correspond to deposit types at banks and thrifts. The correspondence with credit
union terminology is as follows: transactions accounts with credit union share drafts, MMDAs with credit
union MMDAs, and passbook savings with credit union regular shares. Non-interest bearing accounts
have no close parallel in credit unions.
Source: Original papers except OTS.
Table 4 provides estimates of premia.34 Berkovic and Liang (1991) studied premiums paid by
the FDIC for core deposits for failed banks during 1997, finding NOW and MMDA premia of
4.9 percent and 4.0 percent, respectively.35 Hutchison and Pennachi (1996) used Federal
Reserve survey data on commercial banks for estimating deposit rate and balance parameters
in their contingent claims model. They estimated one version of their model which allowed
34
Jarrow and van Deventer (1998) also report that the average premium for all accounts from auctions
of failed bands reported by the RTC was 2.32%. This number includes premia for all types of NMDs.
35
These results were reported for bank and thrift accounts, not for credit union accounts.

deposits to grow and one no-growth version using detrended deposit data. We report the
median results for all banks from their no-growth model. This is the only study that does not
estimate the median MMDA premium to be less than the median transaction account
premium. The Janosi, et al. (1999) one-bank study is an application of the Jarrow and van
Deventer (1998) contingent claims model. They estimated premia for one commercial bank
that undertook mergers during the period of the study. The estimates differ widely depending
on whether they estimated model parameters from total bank deposits, which grew rapidly
during the period, or from “demerged” deposits, which ignored all new deposits from mergers.
The premia from data including mergers are much larger than for demerged data; for example,
for transaction accounts the median premia were 26.2 percent and –6.9 percent, respectively.
They claim that the true premia lie somewhere in between, and we report the average of the
two numbers in the table. Their non-interest-bearing median premium is extraordinarily high;
this could reflect unusual conditions in the one bank. They also estimated the model with
aggregate Federal Reserve data on NOW accounts, which resulted in a much lower premium
of 2.7 percent. McGuire and Newell (1997) studied one commercial bank, Commerce
Bancorp, Inc., of Cherry Hill, New Jersey, using a proprietary model, which, as described in
the article, appears to be similar to the McGuire Performance Systems model. O’Brien’s
(2000) estimates are based on his contingent claims model. Although he reported results for
asymmetric and symmetric deposit rate response to rising and falling market rates, we report
only the symmetric results in the table, simply because they are lower and closer to the
estimates in the other studies. His asymmetric model, for example, estimated transaction
account premia at 21.1 percent and MMDA premia at 12.2 percent. Both sets of O’Brien’s
premia estimates are the largest found in our survey.36 Selvaggio (1996) used the OAS
methodology on one bank, but he applied it only to non-interest-bearing accounts. It is
included in the table merely to show that non-interest-bearing accounts can be expected to
have the largest deposit premia. The OTS reports average deposit premia (and durations)
estimated by its model every quarter. We collected the quarterly numbers from June 1998
through December 1999 and computed the median premia for the table.
A pattern that emerges from Table 4 is that the premia of NMDs varies according to the
degree of interest rate sensitivity of the deposit type. For each study, with the exception of
Hutchison and Pennachi, deposits that are interest rate sensitive, such MMDAs, have lower
36
It should be noted that O’Brien’s premia estimates include the impact of expected growth in deposits.

estimated premia than deposits that do not “reprice” as frequently, such as transaction
accounts. Both McGuire and Newell, a one-bank study, and OTS, which includes many thrifts,
find the highest premia on transaction accounts, followed by passbook savings and then
MMDAs. These are encouraging signs of internal consistency.
It is clear from Table 4 that the type of valuation model used leads to quite different estimates
of premia. The OTS NPVM is a deterministic present value model, with no explicit interest rate
uncertainty. It posits one path of future market rates (in this case the secondary CD rate) and
produces one path of deposit rates and balances. The other methods incorporate multiple
paths of these variables. In other words, the other methods have greater scope for evaluating
the potential exercise of options by the financial institution and its depositors. Another
important difference in methods is whether they allow deposits to grow through new accounts
or additions to old accounts. For consistency with the NEV balance sheet “snapshot”
approach, a “no-growth” estimate of premia would be preferred. Only Hutchison and Pennachi
and OTS explicitly rule out deposit growth. However, the assumption about growth does not
affect the results systematically. Hutchison and Pennachi, who assumed no growth, estimated
higher premia than Janosi et al. (FED), who used aggregate bank data with no detrending for
growth. Apparently, the effect of the growth assumption does not outweigh the other model
differences, such as the way market rates are assumed to evolve, the responses of retention
rates and deposit rates to market rates, the nature of the data, and the time period.
We find that some of these studies are more useful than others in providing estimates of
premia (and durations). First, we do not consider studies of individual banks, because the
studies of many banks (such as O’Brien) indicate that premia on individual banks can vary
widely. This eliminates the one-bank results of Janosi, et al, McGuire and Newell, and
Selvaggio. The O’Brien symmetric model results are then outliers, larger than any other
studies. As noted above, his asymmetric model results were even higher. We decided to
eliminate the O’Brien results. Finally, although we can find not specific problems with the
Hutchison and Pennachi methodology, the MMDA result, with its higher premium than the
transaction account, suggest that these results be treated with caution. The median, or sole,
premia for the remaining studies are the following:
• Transactions accounts (Berkovic and Liang, Janosi (FED) and OTS), 4.9
percent;
• MMDAs (Berkovic and Liang, OTS), 3.0 percent;

• Passbook accounts (OTS), 4.0 percent.
Based on all we have been able to learn from this research survey, we take these results as
“ballpark” premia for NMDs, and below we use the same studies plus one additional study for
duration estimates.
Table 5 provides estimates of duration for various types of NMDs from the same sources
reported in Table 4 and one additional, that of O’Brien, et al (1994).37
Table 5. Duration Estimates for Non-Maturity Deposits

Description: Summarizes published estimated of durations of different types of non-maturity deposits.
The durations are in years.
Period of
Transactio
Model Non-Interest
Author(s) Institution ns MMDA Passbook
Estimatio Bearing
Accounts
n
Berkovic & Failed
Liang (1991) Banks 87 – 90 n.a n.a
Hutchison & Over 200
Pennacchi Banks 86 – 90 6.7 0.4
Janosi et al One
Ind. Bank Bank 88 – 96 5.2 -3.0 6.9
FED
Janosi et al Aggregat
Agg. Banks e Banks 88 – 95 2.4
McGuire &
Newell One
(1997) Bank 88 – 95 n.a n.a n.a
O’Brien 75 – 100
(2000) Banks 83 – 94 1.1 0.5
O’Brien et al
(1994) 2.3 2.9
OTS Thrifts 98 – 01 2.8 1.3 3.4 2.0
Selvaggio One
(1996) Bank 91 – 95 3.0
Note: The account names correspond to deposit types at banks and thrifts. The correspondence with credit
union terminology is as follows: transactions accounts with credit union share drafts, MMDAs with credit
union MMDAs, and passbook savings with credit union regular shares. Non-interest bearing accounts
have no close parallel in credit unions.
Source: Original papers, except OTS.
Like the premia, the estimated durations of NMDs also vary according to the degree of interest
rate sensitivity of the deposit type. Deposits that are interest rate sensitive tend to have
37
Some of these results are also reported in Mays (1997).

shorter durations, such as the MMDA durations, which are approximately one year or less.
They reprice quickly as rates change in a manner similar to floating rate notes, which have low
durations. Passbook savings tend to have the highest durations, followed by transactions
accounts and MMDAs.
Eliminating the same studies as described above, we find the following range of estimates of
NMD durations (in years):
• Transaction accounts:
! Janosi, et al (FED) – 2.4
! O’Brien, et al. – 2.3
! OTS – 2.8
• MMDAs:
! OTS – 1.3;
• Passbook accounts
! O’Brien, et al – 2.9
! OTS – 3.4
We take these durations as reasonable estimates of the interest rate risk of NMD deposits and
useful guidance for the choice of effective maturity. Therefore, we recommend the durations
in Table 6 as the acceptable range of effective maturities for NMD methods that do not specify
NMD cash flows.
Table 6. Recommended Durations for NMDs

Description: These durations are recommended as the final (“effective”) maturities of NMDs for credit
unions that do not explicitly model NMD cash flows. They are acceptable for assigning NMDs to
maturity buckets and for the final maturity when a NMD is treated as a zero-coupon bond.
Min Mid Max

Share Drafts (Transaction
Accounts) 2.0 2.5 2.8
Regular Shares (Passbook) 2.5 3.0 3.5
MMDAs 0.5 1.0 1.5

V. DATA AND ESTIMATION REQUIREMENTS OF DIFFERENT METHODS
Any attempt to evaluate the applicability and suitability of the differing methods of valuing non-
maturity deposits must take into account both the data required as inputs and the analytical
methodology used to convert the inputs into a valuation. Some of the methods summarized
above use relatively simple data inputs, but complex numerical procedures. Others are based
on simpler statistical techniques but require more extensive (in terms of the number and
scope) data inputs.
The implementation of any valuation method necessarily implies careful application of a

generalized model to the circumstances of a specific institution. This in turn requires
understanding the internal logic and assumptions behind the model, and reasoned judgment
and expertise in adapting the model to those specific circumstances. The Office of the
Comptroller of the Currency (OCC) has developed a detailed outline of the steps a financial
institution should follow in developing or applying a risk or valuation model.38 Interested
readers are referred to the complete report; we refer to those portions that we feel are most
applicable to the issues at hand.
Several of the methods summarized above could be loosely described as “time series
methods,” in the sense that they use historical data on individual institutions or market
aggregates to obtain statistical estimates of parameters such as the decay rate on account
balances. These statistical parameters then become the inputs for the main valuation method.
The principal drawback of this type of method is the implicit assumption that parameters
estimated from past data are suitable inputs for a forward-looking valuation method. In other
words, the past is a valid predictor of the future. For institutions that are undergoing major
changes, such as due to merger activity or changes in the local or national economic
environment, this may not be a valid assumption.
Another drawback is that because long series of observations are required in order to obtain
statistically valid parameter estimates, the result is to take the average sensitivity over different
economic conditions. Thus, for example, the sensitivity of deposit balance changes to
changes in aggregate income may be different in periods of economic growth than in a
recession, but this will not be detected. O’Brien’s method explicitly tackles one aspect of this
38
Office of the Comptroller of the Currency, Bulletin 2000-16, “Risk Modeling and Model Validation”.
problem, by making the change in the rate paid on balances be different when market interest
rates are increasing than when they are decreasing.
Table 7 summarizes the essentials of the different methods, in terms of the data inputs used,
and the type and complexity of the estimation methodology. Many of the methods share either
general methodologies or the types of inputs. The general implications of these
methodologies or inputs are considered in section A, while issues that are specific to each
method are discussed in section B.
Table 7. Data Inputs and Methodologies

Description: This table summarizes the various inputs required for each method of valuing NMDs, as
well as the statistical or other methods used to perform the valuation. All items marked with * are
institution-specific. All others are market or macroeconomic.
Method Input Variables Methodology

Simple Methods Deposit Balances* Valued at par, or static
discounted cash flow, using
assumed effective maturity.
Hutchison & Pennacchi Deposit offered rate* Symmetric responses to
(1996) Deposit Balances* changes in interest rates.
Non-interest cost* Equilibrium contingent claims.
Risk-free rate Vasicek term structure model.
Jarrow & van Deventer (1998) Risk-free rate Symmetric response to changes
(Kamakura Corporation) Deposit offered rate* in interest rates.
Deposit balances* Heath-Jarrow-Morton (1 factor)
interest rate model.
Non-interest cost*
Arbitrage-free contingent claims.
McGuire Performance Deposit balances* Simultaneous-equations time-
Solutions Treasury yields series estimation of deposit rate
and balance equations.
Deposit offered rates*
Specific variables and form of
Non-interest cost* equations unique for each
Other institutional factors* institution.
Seasonal factors
O’Brien (2000) Deposit balances* Asymmetric response to
Required reserve ratio changes in market rates.
Deposit offered rate* Cox-Ingersoll-Ross interest rate
model.
Non-interest cost*
Arbitrage-free contingent claims.
Risk-free market rate
Nominal aggregate income

Method Input Variables Methodology
OTS (2001) Non-interest cost* Static discounted cash flow
Deposit balances* Industry-wide estimates of key
Deposit offered rate* parameters.
LIBOR rates
Industry retention rate
Institution’s retention rate*
Interest rate sensitivity*
Selvaggio (1996) LIBOR (1 & 3 Month) Discounted cash flow with OAS.
Average cost of DDA balances Cox-Ingersoll-Ross interest rate
Deposit balances* model and simulations.
Deposit offered rate*
Reserve requirement
Spot Treasury rate
Nominal aggregate income
Source: Original papers.
A. General Considerations
There are some important issues that are common to several of the methods discussed in
Chapter II, and that need to be addressed so that the NCUA and its examiners can be assured
that any method that is adopted by member credit unions will result in consistent valuations
and treatment of NMDs. For example, all of the advanced methods involve estimating
parameters that relate the rate paid on deposits and deposit balances to market and
institution-specific variables. If there is not uniformity in how those parameters are estimated
and applied, there is the potential for major discrepancies between valuation results of
different credit unions.
The issues common to many of the methods can be grouped into those that relate to
considerations concerning the data that is used to estimate the parameters and value the
deposits, and the estimation of the parameters themselves.
1. Data Considerations
Most of the methods of valuing the economic value of NMDs use some form of statistical
analysis to infer the parameter values for use in the valuation. The econometric issues
involved with this are considered in the next section, but it is also important to understand the
data requirements such methods involve.

It is possible that a valuation method is conceptually sound and has been programmed
correctly. However, if there are inconsistencies or errors in the data that are used as inputs,
the resulting valuation and any other outputs will be useless. Put another way, having a good
software program is not a guarantee that the valuations will be accurate. Thus, it is important
that a careful and formal audit of data inputs become part of every valuation performed by a
credit union.
The data used in valuing NMDs come from two sources: internal and external (usually market
related). For data that is internal to the credit union, the normal internal audit functions of the
credit union should be sufficient to ensure that the information used as inputs are an accurate
representation of the credit union’s condition at the time of valuation. For external data, the
credit union’s principal source can be checked against multiple sources. For both types of
data, it is wise to incorporate some procedures to spot errors that may occur, such as
automated filters and visual inspection of the data by experienced personnel.
Another type of data problem that arises is ensuring that all the data are a uniform and
consistent representation of conditions at a single point of time. For example, services such
as Bloomberg and Reuters provide several different LIBOR yield curves; these vary by time of
day (London close, New York close, etc.) or by source (specific banks, etc.). It is important
that a credit union obtaining such external data consistently use the same version and source
to avoid discrepancies.
2. Estimation Considerations
For any method adopted by credit unions, differences in methodologies for estimating
parameters can cause major differences in valuation between institutions. Chapter VI
illustrates this by exploring the sensitivity of an illustrative method is to changes in the input
parameters.
There are several areas where uniformity is important, and where lack of uniformity can create
problems. They are:
a. Length of sample period
As indicated above, all of the time-series and option-based methods involve the use of
parameters that relate the rate paid on NMDs as well as the future NMD balances to various
market and institution-specific variables. These parameters are typically obtained from
regressions between these variables using time-series of past observations. The length of the

sample period that is used to run the regressions can have an impact on the resulting
parameter estimates. There is a trade-off between using more versus less data. The more
observations that are used (the longer the sample period) the more statistically reliable the
estimates will be. However, if the relationships being estimated are not constant but tend to
change through time or in response to different economic situations, a longer sample period
may lead to a parameter estimate that is a blend of the parameter from two different economic
regimes.
For example, if the relationship between NMD offered rates and market interest rates is
different when market rates are increasing from when rates are decreasing, then running the
regression using a sample period that includes both scenarios will lead to a parameter
estimate that is a blended average of the two “true” values of the parameter.
It is therefore important to determine the appropriate uniform length of sample period that
should be used by any institution adopting a particular method of valuing NMDs. Five years of
monthly data will provide sixty observations; this should be sufficient for most regression
methods. More observations are advisable for some statistical methods (such as some of the
time-series and simultaneous equations methods), and even simple regression methods such
as ordinary least squares (OLS) can use more observations if there are many parameters to
estimate in a single regression. The issue of whether a particular five-year period includes
observations from more than one economic regime is an issue that must be taken into account
at the time of estimation.
b. Estimation method used
If individual credit unions are responsible for estimating and applying a specific method, rather
than relying on commercially provided inputs and parameter estimates, uniform methods of
estimating the parameters will aid compatibility. If one institution estimates a set of
parameters using simple OLS regression with no correction for errors, such as
heteroscedasticity or serial correlation, the resulting parameter estimates may be very different
from those that are obtained by another institution correcting errors using more sophisticated
regression techniques.
For the most part, these issues are addressed in each published model. This means,
however, that any institution adopting a particular model needs to understand all of the
statistical and econometric issues involved in using it.

c. Frequency of updating parameter estimates
How often should a set of parameters that are used in a valuation method be re-estimated and
updated? The answer in part is an empirical one, since it depends on the extent to which
relationships change in response to different economic regimes. It also depends on the
frequency with which the valuations are performed. While NCUA examiners typically might
review such valuations annually, large sophisticated credit unions may want to value their
NMDs more frequently as part of an active asset-liability management program.
From the point of view of NCUA and its examiners, updating of parameter estimates in a
uniform manner across institutions would aid comparisons of the riskiness and NEV of
different institutions, and would provide a check on the reasonableness of individual
valuations. Ideally, parameter estimates should be based on not only common lengths of
sample periods, but sample periods that end on the same date. To see why this is important,
consider the example of an examiner performing a review of two credit unions, ABC and XYZ,
with similar exposures to market variables and similar characteristics. All else equal, these
two institutions should have similar parameter estimates and therefore similar NMD valuations.
Assume also that parameter estimates are made with 5-year sample periods of monthly data
that end the month before the examination takes place. If institution ABC is examined in
January and institution XYZ is examined in the following December, there will be only 48 data
points in common in the two samples used to obtain the parameter estimates. If the earliest
data that is used in the case of ABC but not XYZ reflects a period of rising interest rates and
economic growth, while the latest period is one of falling interest rates the resulting
regressions may yield very different parameter estimates.
In order to minimize the fluctuations in valuation that might occur due to this problem, it is
suggested that parameters be updated each time a valuation is performed, using data up to
the date of valuation. Since the valuation date for reporting to NCUA will typically be the end
of a quarter, this will ensure that parameter estimates across institutions reporting at the same
time will reflect as close to a common sample and data as possible.
B. Method-Specific Considerations
1. Static Discounted Cash Flow Methods
The method of valuing NMDs adopted by the Office of Thrift Supervision uses a combination
of market data, industry-wide parameter estimates, and institution-specific data and

parameters. The valuation is a three phase process: first the expected future path of interest
rates is obtained from the current yield curve; second, based on industry-wide parameters, the
future rates paid on deposits is obtained from the expected future path of market interest rates;
finally, based on both industry-wide and institution-specific parameters the future level of
deposit balances for each of the next 360 months is calculated. When all these steps have
been followed, the cash flows attributable to the deposits are then discounted to complete the
valuation process.
There are two aspects to the OTS NPVM. One is the valuation method, involving the static
discounted cash flow approach just discussed. The other is the regulatory policy it illustrates.
By adopting a “top-down” approach, the OTS is removing from the hands of thrift managers
the issue of how to value NMDs by making a unilateral decision for all thrifts. From the point of
view of an individual institution, the advantage of the OTS NPVM is that the thrift does not
have to worry about any of the issues of estimating and updating parameters discussed in the
previous pages. The OTS does this for all member thrifts, and the updated parameter
estimates are made available on its website.
From the NCUA’s perspective, however, the issues remain. Whether or not the NCUA should
put itself in the position of providing statistical parameter estimates on a regular basis, and
requiring all insured credit unions to adopt the same valuation method is a policy decision that
is beyond the scope of this study.
2. Contingent Claims
All of the contingent claims methods involve the consideration of many possible interest rate
paths. This is because the option-related features of NMDs mean that future cash flows are
not known with certainty, but depend on the particular path that interest rates (both market
rates and deposit rates) take. This requires knowledge of both interest rate modeling and
simulation. Implementing an interest rate model in a spreadsheet or other software program is
not trivial, and is prone to many possible sources of error. Adding the simulation component
only compounds the problem, since even with commercially available simulation packages the
simulation of a complex interest rate process is not necessarily easy to implement.
3. Option-Adjusted Spreads
Many securities are valued using OAS methods. As described above, this method involves
finding the OAS that must be added to the discount rate in order that the present value of the

expected future cash flows (which vary with the interest rate scenario, due to the embedded
options) on average equate to the observed market value of the asset. This method is widely
used to value securities such as mortgage-backed securities (MBSs), collateralized mortgage
obligations (CMOs), and other types of bonds that contain embedded options.
The drawback of this approach as it relates to valuing the NMDs of credit unions is that there
is not an active market for core deposits of credit unions. Therefore, one of the key inputs to
the approach (the market observed premium) is not generally available. Instead, an assumed
OAS must be added in order to arrive at the estimated value. However, how should a credit
union go about estimating an appropriate OAS for valuing its NMDs? It is well known by
market participants that the OAS of MBSs and CMOs, for which there are quite active markets,
vary with changes in interest rates, economic activity, and changes in investor sentiment.
There is no model of OAS for credit union deposits that is known to the authors of this report.
In the absence of such a model, any attempt to use OAS methodology to value NMDs will be
subject to potentially severe risk of mis-valuation. To the extent, however, that a proxy for the
missing variable (observed deposit premia) can be found, this approach may be adopted with
caution.
C. Conclusions
We have considered some of the practical aspects involved in adopting some of the methods
described in Chapter II. Many of the methods are not only complex and sophisticated in their
approach to tackling the problem of valuing NMDs; they also require careful and sophisticated
application and the availability of a wide range of internal and external data.
We must keep in mind that many credit unions are small organizations, and many have limited
experience or expertise in the arena of financial modeling. All these factors tend toward a
conclusion that most credit unions will be quite adequately served by using simple methods for
valuing their NMDs, as long as they are used in a careful manner, consistent with the specific
circumstances of their institution. The key question then becomes at what point does a credit
union become too large or too sophisticated to use simple valuation methods.

VI. AN ILLUSTRATIVE SIMPLE MODEL
As explained in the Introduction, the reason for constructing this model was to illustrate an
example of a present value model with NMD cash flows that credit unions might use in lieu of
the more complex models. We were guided by several considerations:
(1) Even a simple model should include cash flows for deposit retention, deposit
interest and servicing costs.
(2) Deposit retention could be based on data from OTS, since it is the most recent data
available for institutions similar to credit unions.
(3) The deposit premium should be based on the present value of cash flows for
consistency with the calculation of NEV.
(4) The discount rate should be the Treasury rate based on the assumption that the
deposit has no credit risk.
(5) The durations produced by the model should be similar to the durations
recommended in Table 6, which are based on durations estimated in the literature.
(6) Deposit premia produced by the model should be similar to those estimated in the
literature, as summarized in Section IV.
We adopted a simple model that meets these criteria. The main features of the model are: (1)
the monthly Treasury spot rate stays constant at its value on the NMD valuation date. (2) The
deposit rate stays at a specified, constant percentage of the Treasury rate. (3) The deposit
decays at a constant rate until a specified final maturity at which time the remaining deposit is
withdrawn.39 There is no question that this model is very simple, and some credit unions may
find it too simple for their needs. For those credit unions which up to now have used only a
final maturity assumption with no specification of cash flows, however, even using this simple
model can be an important step toward more accurately valuing NMDs, incorporating NMDs in
the NEV calculation in a consistent manner, and focusing management attention on the issue
of understanding deposit rate and balance behavior.
39
A model can be specified in terms of a retention rate or a decay rate. One implies the other. We find
it convenient to specify a decay rate. For example, a 12 percent decay rate, expressed as a monthly-
compounded rate, means that 1 percent of the balance is withdrawn each month. The corresponding
annual retention rate is approximately 88 percent.
For the deposit decay rate in our illustrative model, we consulted the OTS NPVM
documentation, which includes the values of the deposit retention parameter for the four
categories of NMDs: transaction accounts, non-interest bearing accounts, passbook savings
accounts, and money market deposit accounts. These can be converted into annualized
decay rates as shown in Table 8. They are useful as an indication of how decay rates vary
across types of deposit, and how they have changed in the five years since the OTS first
began tracking these figures
Table 8. Industry-Wide Deposit Decay Rates for Thrift Deposits

Description: The OTS NPVM uses as a key input a parameter that reflects industry-wide median decay
rates for all thrifts reporting new account balance information, and for each deposit type. The following
table reports the annualized decay rates for the 1994 version of the model, and the recent version dated
April 2001.
Deposit Type 1994 OTS NPV Model 2001 OTS Updated Model
Transaction Accounts 20.24% 22.70%
Money Market Accounts 43.94% 37.74%
Passbook Accounts 15.94% 21.54%
Non-Interest Bearing Deposits 43.94% 42.08%
Source: OTS (1994), OTS (2001).
Note that MMDAs have by far the highest decay rate of all interest-bearing account types, very
similar to that of non-interest bearing accounts, in keeping with the fact that they are more
interest-rate sensitive. The decay rate of transaction accounts and passbook accounts has
increased in the seven years since 1994; while for MMDAs and non-interest bearing accounts
it has decreased. This is an example of the sensitivity of model parameters to changes in
economic conditions and interest rate scenarios.
For the relationship of the deposit rate to the Treasury rate, we reviewed the ratio of historical
credit union NMD rates to the 3-month constant maturity Treasury rate.40 Servicing costs were
based on OTS data. The effective maturities were then selected to produce durations similar
to those in Table 6 and deposit premia similar to those studies considered most useful as
discussed in Section IV.
40
Sources: www.cuna.org/data/cu/research/cu_stats and www.federalreserve.gov/releases/HTS/data

A. Simple Valuation Results
The present value of the cash flows associated with each deposit type was calculated
according to the formula
T
(i + c )Dt −1 + (Dt −1 − Dt )
PV = ∑ . (11)
t =1 (1 + rt )t
where r is the Treasury rate, i is the interest rate paid to depositors, c is the cost of acquiring
and servicing the deposit, and D is the deposit balance. The decay rate, d, determines each
period’s balance according to the equation
Dt = Dt −1 (1 − d ) (12)
The length of each period in the model is one month. The valuations were performed first for
the base case scenario and then after a shock to interest rates of +/-300 basis points, similar
to the procedure used in NEV calculations. When the yield curve is shocked, adjustments
must also be made to both the deposit rates and the decay rates. Changes in the deposit
rates were made by assuming that the deposit rate stays at the same percentage of the
market rate; thus, the change in deposit rate is not one-for-one with the change in market rate.
The decay rate change reflects the interest rate sensitivity of NMDs: because of the way
deposit rates change relative to market rates, as market rates increase (decrease) the spread
between the market and deposit rates increases (decreases). This then results in customers
being more (less) likely to withdraw their funds, so the decay rate increases (decreases).

Table 9. Base Case Valuations and Results from +/- 300 bp Shocks
Description: Based on a simple present value model of NMD cash flows. The user-defined inputs are
constant decay rates, the constant ratio of the deposit rate to the market rate, the non-interest cost rate,
and the maturity. The market rate is the 3-month constant-maturity Treasury rate.
Shock (bps)
-300 0 +300
Share Draft Premium -0.68% 4.64% 8.98%
Average Life 2.92 yrs 2.72 yrs 2.59 yrs
Duration 2.84 yrs 2.57 yrs 2.36 yrs
Deposit rate 0.30% 1.20% 2.10%
Market Rate 1.00% 4.00% 7.00%
Maturity 4 yrs 4 yrs 4 yrs
Decay rate 17% 21% 24%
Non-int. cost 1.35% 1.35% 1.35%
Regular Premium 2.35% 4.50% 6.24%
Shares
Deposit rate 0.70% 2.80% 4.90%
Market Rate 1.00% 4.00% 7.00%
Maturity 4.5 yrs 4.5 yrs 4.5 yrs
Non-int. cost 0.02% 0.02% 0.02%
MMDA Premium -0.32% 0.75% 1.70%
Deposit rate 0.74% 2.95% 5.16%
Market Rate 1.00% 4.00% 7.00%
Non-int. cost 0.65% 0.65% 0.65%
The results of these valuations for the base case and +/- 300 basis points are reported in
Table 9. The model’s results are consistent with our reading of the research. Share drafts
have higher premia but shorter durations than regular shares. MMDAs have the lowest premia
and durations. Shocking the market rate does not lead to drastic changes in the valuations.
The occurrence of negative premia is a useful reminder that this can occur for actual credit
unions under some interest rate scenarios. We conclude that simple present value models
can produce premium estimates that are reasonable and credible. The premia and durations
reported are valid only for the deposit rates, market rates, decay rates and effectives
maturities used in this specific example. Other combinations of these variables will lead to
different premium and duration results. This table is not to be interpreted as the definitive
valuation of credit union deposits using a simple valuation model.

An example may help explain how the base case and shocked premiums may be used in the
context of asset-liability management. Consider a simple credit union that has $100 (market
value) in assets, all invested in mortgage-backed securities (MBS). Those securities are
funded by a deposit with a face value of $100. If the deposit premium is 5%, then the liability
value of the deposit is $95, and the credit union’s net capital is therefore $5. Suppose there is
a shift in interest rates of +300 basis points, and the value of the MBS falls by 20% to $80,
while the deposit premium rises to 10%. The market value of the deposit therefore falls from
$95 to $90, while the net capital falls by the decrease in asset value ($20) less the increase in
deposit premium ($5), to -$10. In other words, the impact of the deposit premium, in a NEV
context, is not in the level of the premium but in the change in premium due to the change in
interest rates.
B. Sensitivity Analysis
To determine how changing the assumptions can impact the model results, we repeated the
calculations used to construct Table 9, varying in turn key variables. The aspects of the model
that we chose to examine were: the assumed maturity, the assumed decay rate, the
assumption of a parallel shift in the yield curve, and the spread between the deposit rate and
the market rate.
A key variable in the simple present value model is the assumed maturity, which was selected
to make the model produce durations similar to those recommended in Table 6. It is of
interest, therefore, to investigate the sensitivity of durations and deposit premia to the choice
of maturity. The following table compares the base case results for each type of NMD, and
results for shorter and longer maturities.

Table 10. Sensitivity of Base Case Valuations to Assumed Maturity
Description: The impact on base case valuations to changing the assumed maturity. All remaining
assumptions are the same as in Table 9.
Assumed Maturity
Shorter Base Case Longer
Share Draft Premium 2.56% 4.64% 6.05%
Deposit rate 1.20% 1.20% 1.20%
Market Rate 4.00% 4.00% 4.00%
Non-int. cost 1.35% 1.35% 1.35%
Shares
Deposit rate 2.80% 2.80% 2.80%
Market Rate 4.00% 4.00% 4.00%
Maturity 3 yrs 4.5 yrs 7 yrs
Non-int. cost 0.02% 0.02% 0.02%
MMDA Premium 0.42% 0.75% 1.13%
Deposit rate 2.95% 2.95% 2.95%
Market Rate 4.00% 4.00% 4.00%
Maturity 1 yr 2 yrs 3 yrs
Non-int. cost 0.65% 0.65% 0.65%
Using longer maturities does not produce a large change in calculated premiums.
Furthermore, the resulting durations remain roughly consistent with the empirical evidence.
The second sensitivity analysis explores the decay rate. The results for share drafts and
regular shares are reported in Figure 4 and Figure 5. Note that the premium for Regular
Shares declines at a slightly faster rate than that of Share Drafts as the assumed decay rate
increases. The premia for both share drafts and regular shares are relatively insensitive to the
assumed decay rate: the total variation over the 10% range in decay rates is only about 1% in
premium value. Thus, we conclude that the choice of decay rate, by itself, is not a crucial
assumption.

————————————————————————————————————————————
Description: The accompanying figure shows how the premia vary if the decay rate is allowed to vary.
————————————————————————————————————————————
5.50%
5.25%
Deposit Premium
5.00%
4.75%
4.50%
4.25%
4.00%
15% 16% 17% 18% 19% 20% 21% 22% 23% 24% 25%
Assumed Decay Rate (Annualized)
Share Drafts Reg. Shares
Figure 4. Sensitivity of Deposit Premiums to Assumed Decay Rate
There are two similarities in the effect of changing the decay rate on the estimated duration
with the impact on the estimated premia: the magnitude of the change is not very great,
approximately +/- 0.5 years for share drafts and regular shares; and the duration of regular
shares decreases faster than that of share drafts as the decay rate increases.
————————————————————————————————————————————
Description: The accompanying figure shows how durations vary if the decay rate is allowed to vary.
————————————————————————————————————————————
3.50
3.25
Duration (Years)
3.00
2.75
2.50
2.25
2.00
15% 16% 17% 18% 19% 20% 21% 22% 23% 24% 25%
Assumed Decay Rate (Annualized)
Figure 5. Sensitivity of Deposit Durations to Assumed Decay Rate

Next, we consider non-parallel shifts in the yield curve. Interest rate sensitivity analysis
performed for NEV (or the OTS NVPM) assumes that changes in the yield curve take place in
a parallel fashion, with the entire curve moving up or down by the same number of basis
points. In reality, the yield curve rarely shifts in a parallel fashion, and the short end of the
yield curve is much more volatile than the long end. We therefore consider the impact on the
base case valuations from Table 9 of two non-parallel shifts in the yield curve: a lowering and
steepening of the curve, and a curve inversion when rates rise.
The following figure shows the three yield curves used: the Treasury spot rate curve as it was
on March 31, 2001; the spot rate curve when it is lower and steeper; and the spot rate curve
when it is higher and inverted.
————————————————————————————————————————————
Description: Treasury spot rate (zero-coupon) curve as of March 31, 2001, taken from the Selected
Asset and Liability Price Tables (OTS, 2001), as well as curves resulting from: i) a lowering and
steepening; and ii) a raising and inverting of the entire yield curve.
————————————————————————————————————————————
8%
7%
6%
5%
4%
3%
2%
1%
0%
1 11 21 31 41 51 61 71 81
Months to Maturity
Note: The yield curve reported by OTS contained spot rates for maturities of 3, 6, 12, 18, 24, months, etc. Linear
interpolation was used to obtain spot rates for intermediate maturities.
Figure 6. Yield Curves Used in the Non-Parallel Shift Sensitivity Analysis
The analysis in Table 9 is then repeated, except that the yield curves shown in the above
figure are used instead of the parallel shift of +/- 300 basis points. The results are reported in
Table 11.

Table 11. The Effect of Non-Parallel Shifts in the Yield Curve
Description: The impact on base case valuations of non-parallel shifts in the yield curve instead of +/-
300 basis points. All remaining assumptions are the same as in Table 9.
Shock (bps)
Lower & Base Higher &
Steeper Case Inverted
Share Draft Premium 1.49% 4.64% 6.81%
Deposit rate 0.32% 1.20% 2.11%
Market Rate 1.06% 4.00% 7.02%
Non-int. cost 1.35% 1.35% 1.35%
Shares
Deposit rate 0.74% 2.80% 4.91%
Market Rate 1.06% 4.00% 7.02%
Maturity 4.5 yrs 4.5 yrs 4.5 yrs
Non-int. cost 0.02% 0.02% 0.02%
MMDA Premium 0.26% 0.75% 1.11%
Deposit rate 0.78% 2.95% 5.18%
Market Rate 1.06% 4.00% 7.02%
Non-int. cost 0.65% 0.65% 0.65%
The results are similar to those of the parallel shift in rates (Table 10) with the exception of the
case of regular shares. The premium on regular shares increases when the yield curve is
lower and steeper and decreases when the yield curve is higher and inverted. This apparently
anomalous behavior is due to the interaction between the discount rate, the deposit rate and
the decay rate. We would expect that an upward yield curve shift would reduce the value of
the deposit due to the higher discount rate, thus increasing the deposit premium (which is par
less the value of the deposit). This effect can be offset, however, by two other factors: (1) A
higher deposit rate which increases deposit value and decreases deposit premium; and (2) A
higher decay rate, which increases deposit value because cash flows are paid out sooner and
decreases the premium. In the case of regular shares for this interest rate scenario, these two
effects overwhelm the effect of the higher discount rate. This is undoubtedly due to the fact

that in this scenario, the longer term discount rates are not shifting up by the full 300 basis
points shown in Figure 6. A similar explanation applies in the case of the falling and
steepening yield curve.
The impact of the spread between the deposit rate and the market rate is shown explicitly in
the next sensitivity analysis. We focus on this element by holding constant the market rate
(zero-coupon Treasury curve) and varying the deposit rate while holding constant all other
inputs. The results may be seen in Figure Figure 7.
————————————————————————————————————————————
Description: Premia variation when the deposit rate is allowed to vary, thus affecting the spread
between the market rate and deposit rate.
————————————————————————————————————————————
14%
12%
10%
Deposit Premium
8%
6%
4%
2%
0%
-2%
-4%
0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% 3.5% 4.0%
Deposit Rate
Figure 7. Sensitivity of Deposit Premia to Spread between Deposit Rate and Market Rate
The premia of both share drafts and regular shares are very sensitive to changes in the spread
between the deposit rate and the market rate. If this spread is less than the non-interest
service charges associated with the NMDs, then the premium will be negative. The lower the
deposit rate the higher the premium, all else equal.

VII. RECOMMENDATIONS
NERA’s recommendations to NCUA are presented in this section. We again stress that NERA
has no prior preference for any particular valuation approach. We have endeavored to make
recommendations that are purely in the best interests of NCUA and its member institutions.
Our recommendations fall into three categories: those that relate to the assumptions on
effective maturity needed for some vendor methods; those that relate to how NCUA can
improve the understanding of how credit union deposits behave differently from those of banks
or thrifts; and those that relate to specific methods.
A. Recommendations on Effective Maturity Assumptions
One of the specific questions we were asked to address is, for those methods that do not
explicitly model the cash flows associated with NMDs but instead assume an effective
maturity, what is a reasonable assumption? We have argued that the answer to this question
is to use durations similar to those estimated in the literature we surveyed. These are
presented in Table 6 (as well as Table 1).
The ranges of durations recommended here reflect the durations of NMDs estimated in the
literature and by OTS, and are intended only for those credit unions that do not model NMD
cash flows. Credit unions that use fully-specified methods to value their deposits, such as
OTS NPVM and contingent claims methods, should use the final maturity specified in each
model.
B. Suggestions on How to Assess the Accuracy of the Recommended Effective

Maturities for Credit Union Deposits
It is apparent from all of the literature and evidence from practitioners in the field of valuing
NMDs that the most important aspect of any valuation method is the experience of the specific
institution. As we have stated on numerous occasions in this report, there is little or no
published data on the sensitivity of credit union balances to changes in interest rates, nor is
there evidence on the retention rate of credit union balances. Thus, the durations on which
the recommended maturities for use in simple valuation methods are based do not reflect the
experience of credit unions’ deposits, but those of banks and thrifts.
We feel that NCUA would be providing a major service to the credit union industry by
undertaking to estimate the retention rate of credit union deposits as a step toward better
estimates of their durations. There are two ways that NCUA could perform such an empirical
estimation.
1. Modify the Call Report
By making some minor modifications to the call report, NCUA would be able to calculate
median retention rates in the same manner as OTS. This would involve asking credit unions
to keep track of and report not only the aggregate balances of deposits, but also the amount of
new accounts of that type opened that quarter. The retention rate for a given deposit type in a
given quarter is calculated using a four step process:41
1) Calculate the retention rate for the current quarter using the following formula:42
Current Aggregate Balances - Aggregate Balances in New Accounts

Retention rate = (12)
Previous Quarter' s Aggregate Balances
2) Repeat step 1 for each of the previous two quarters.
3) The median of the three rates calculated in steps 1) and 2) is chosen by

discarding the highest and lowest rates.
4) Convert the median retention rate from step 3) to an annualized rate.
While this procedure is not the most exhaustive analysis of retention rates, it has some
important advantages. First, it is relatively easy and straightforward to estimate, without
requiring extensive statistical analysis or sophisticated methodology. Second, it will allow both
credit union management and NCUA to track retention rate behavior in an ongoing fashion.
This will provide significantly more information than is currently available, allowing early
detection of changes in trends, and other unexpected activity.
2. Commission a Statistical Study of Credit Union Deposits
The second approach involves performing detailed statistical analysis on the behavior of credit
union NMDs over lengthy time series. This would be directly comparable to the studies of
bank and thrift data found in the literature surveyed. Such a study would involve collecting
41
See OTS (1994) or OTS (2001), Section 6.D.
42
We are interested in estimating the retention of balances from the previous reporting period excluding
additions to existing accounts. Note that the term “Current Aggregate Balances” will include such
growth, and thus may overstate the “no-growth” retention rate for NEV purposes.

data on a large sample of credit unions, and performing careful statistical analysis of long time
series of aggregate credit union deposit balances.43 It is important to control for the growth in
existing and new accounts in the context of NEV to the extent permitted by the data.
There are two possible sources of data for such a study. The first is the aggregate deposit
balance data on all credit unions filing quarterly call reports with NCUA. The second is to
obtain data from a sample of credit unions on aggregate deposit balances, and a sample of
individual account balances. The call report data is already available in machine-readable
form and in a central location, so the data collection task would be simplified. However, by
using only aggregate balance data, important insight into deposit behavior may be lost which
would be acquired by using samples of individual accounts.44
3. Summary of Retention Rate Recommendations
The first approach described above would enable NCUA to provide ongoing estimates to the
entire credit union industry of the recent behavior of deposits. These numbers could be
reported for the entire industry, as well as by different sized institutions, by region, etc. While it
is a simple approach that gives only an approximation to an institution’s retention rate, it does
yield insights that are not generally available absent such calculations. A more complete
picture of a credit union’s retention rate might be obtained by relating these data to the income
and interest rate environment.
The second approach, the detailed statistical study, would lead to a report that would help
credit union managers and regulators understand how balances in this industry behave in
different economic and interest rate environments. Since so much research has been
published on this topic in the area of banks and thrifts, such a study would allow direct
comparisons to be made where none have been possible before. This in turn would further
aid in determining which valuation methods are most appropriate for credit unions as opposed
to banks or thrifts. It would also help determine whether the range of durations for the different
deposit types that are based on empirical studies of bank and thrift deposits are an accurate
43
A detailed description of the statistical methodology involved in such a study is beyond the scope of
this report. However, the interested reader is referred to the technical descriptions contained in
Hutchison and Pennacchi (1996), Janosi, Jarrow and Zullo (1999), Kahn, Pennacchi, and Sopranzetti
(1999), O’Brien et al. (1994) and O’Brien (2000).
44
For example, the aggregate balance data available in the call reports do not show the extent of
transfers between different deposit types at the same institution (such as from MMDA to transactions
accounts). This might be revealed in analyzing the individual accounts.

reflection of credit unions or need to be modified. Such a detailed statistical analysis would
yield a picture of credit union deposit behavior in various interest rate and economic scenarios,
as well as any differences between large and small credit unions.
C. Recommendations on NMD Valuation Methods for Credit Unions
Finally, we have the following specific recommendations concerning whether the various
valuation methods discussed herein are appropriate for all or some credit unions:
1. Simple vendor methods that do not explicitly model the cash flows of non-maturity
deposits, but instead use assumed effective maturities, are appropriate for small credit
unions as long as the assumed maturities are durations within the recommended range
given in Table 6.
2. There should be uniformity in the choice of discount rates for valuation purposes. Many
methods that were developed for use in banks and thrifts use discount rates that
represent the cost of alternative sources of funds for such institutions (such as LIBOR,
or the FHLB borrowing rate). These rates are not appropriate for discounting by credit
unions unless it is assumed that credit union NMDs have the same default risk as
these sources of funds. If credit union NMDs are default free, the Treasury spot (zero-
coupon) rate curve should be used.
3. Option-adjusted spread (OAS) models are generally not appropriate for credit unions.
Unlike bank deposits, there is no market for credit union core deposits, which makes a
key input to the OAS method (the observed market value of the deposits) unavailable.
Instead, either an assumed OAS must be used, with obvious drawbacks, or deposit
values or OAS from bank and thrift deposits are used. Since credit union deposits may
behave in materially different fashion from those of banks and thrifts, this makes the
accurate and consistent estimate of credit union deposit values using this method
unlikely. Some vendors of this type of valuation method use a proxy for the
unobserved market premium. To the extent that the proxy is reasonable and valid,
such a method may be appropriate for credit unions, as long as they understand the
implications of such a simplifying assumption.
4. Present value methods with relatively complete specifications of NMD cash flows,
including in some cases scenarios of future interest rates, are appropriate for credit

unions with sufficiently detailed history of deposit behavior and sufficient sophistication
to implement them.
5. Contingent claims-based methods (not to be confused with OAS-based methods) are

suitable only for large credit unions that have a detailed history of NMD balances, that
have the requisite understanding of contingent claims methods, and that can
demonstrate that such a method is appropriate for their institution because actual
deposit balance and rate behavior conforms to the assumptions of the models.
6. Valuation methods that incorporate expected growth of deposit balances, while

theoretically sound, are not consistent with the NEV model that is the wider framework
within which NMD valuation should fit.
Finally, at what point should a credit union consider moving from the simpler valuation
methods to ones that are more complex? This is not a question with a simple answer suitable
for every credit union. The answer depends in part on the complexity of the credit union, the
range of assets in which it has invested members’ funds, and the behavior of its deposits. To
the extent that the credit union is investing a significant portion of its balance sheet in
mortgages, or other loans or securities with embedded options, it should consider adopting a
more complex method of valuing its deposits. This is because such assets have very different
interest rate risk than other possible investments; this in turn means that the valuation of the
deposits must be undertaken in a more comprehensive manner to determine the institutions’
interest rate risk.

VIII. IMPLICATIONS AND GUIDELINES FOR NCUA
The recommendations in the previous chapter lay out some important principals and
guidelines. The guidelines can be formulated as a set of possible questions that NCUA should
be seek to have answered when conducting a review of a credit union. They can be grouped
into those questions that are general in nature, and those that relate to specific methods and
approaches to valuing NMDs.
The responsibility of examiners, in reviewing the valuations performed by credit unions of their
NMDs, is to ensure they have a clear understanding of which method is used to value the
deposits, the assumptions made to implement the specified method, and the implications that
follow from a particular set of assumptions.
NCUA and its examiners should always be aware that some credit unions and some vendor
models value NMDs at par. From a regulatory standpoint, this is a conservative approach,
and may be acceptable. However, where credit unions do have good data available on their
deposits over time, they should use that data to assess more accurately how their deposits are
likely to behave in different economic and interest rate scenarios.
In each case, the suggested question is given in bold face type, with a brief discussion
following.
A. General Questions
• Do the characteristics of the deposit match the stated type?
This question is designed to ensure that each deposit type is valued using the method
appropriate for it. If a deposit is incorrectly specified, so that it is valued as a different
type, it could lead to significant mis-valuations and estimates of the interest rate risk of
the credit union.
For example, several of the methods discussed in this report have different versions of
their model, with different parameters, etc., for MMDA accounts and transactions
accounts. An account labeled “Money Market Share Account” might sound as if it is
like any other MMDA, but if the terms of the account call for the deposit rate to be
changed only once or twice a year, this is closer in spirit to a passbook savings
account. A true MMDA-type account would have the deposit rate adjust much more
frequently.

• Does the valuation method used seem appropriate for the credit union?
While many smaller credit unions may find it appropriate to use simple valuation
techniques, there comes a time in the growth of a credit union that it needs to use a
more sophisticated method to value its NMDs. Credit unions that have access to
detailed histories of their deposit behavior and financial modeling capability should take
advantage of advances in NMD valuation methods.
Conversely, credit unions may be tempted to apply a more sophisticated method than
the size and complexity of their institution warrants. Examiners should use their best
judgment to determine that the method used is appropriate.
B. Questions Specific to Particular Methods
1. Simple vendor models
• Does the effective (final) maturity conform to the recommended range for each
type of deposit?
The recommended range of maturities (which are durations) was developed in the light
of available empirical evidence on deposit behavior. Credit unions should only exceed
the maximum recommended maturity if they can demonstrate that a longer duration is
appropriate for their institution.
• If the assumed maturity does not conform to the recommended range for any
deposit type: a) Does the credit union have at least 5 years of monthly deposit
balance data; and b) Using this data, can the credit union demonstrate that the
duration of their deposits is similar to NMD durations estimated in the research
literature?
This question is designed to ensure that a credit union only departs from the
recommended range of maturities when it has valid evidence for doing so. At least five
years of monthly data is required so that any statistical tests performed are valid, and
so that the data is likely to cover more than the most recent interest rate environment.
• Does the credit union use the appropriate discount rate?
This is the Treasury spot rate (zero-coupon) yield curve.

2. Present Value Methods
• Does the credit union have at least 5 years of monthly deposit balance data with
which it can estimate the necessary parameters?
The five-year requirement is to ensure that parameter estimates used in these

sophisticated methods have a reasonable degree of statistical accuracy. The
parameters of a model determine the sensitivity of deposit rates and deposit balances
to changes in market rates and other factors. If these parameter values are not
updated by using the most recently available data to perform the regressions, the
deposits will not be valued inaccurately.
• Market data is typically required as inputs for these models. Does the credit
union have access to a reliable and timely source of such information?
These more sophisticated methods of valuing NMDs will only yield accurate estimates
of the value of the deposits if the parameters estimates are based on accurate data.
Reliable sources for market data include commercial services such as Bloomberg, and
Reuters. OTS publishes the Treasury spot rate (zero-coupon) yield curve prevailing at
the end of each quarter in its publication Selected Asset and Liability Price Tables.
• Were the parameters updated at the time of the most recent valuation, using up-
to-date data?
Since many of the parameters used in these sophisticated methods may vary in
different economic and interest rate scenarios, it is important that each time the
valuation is performed the parameters reflect the most recent and up-to-date
information possible. This includes both market data and data specific to the credit
union.
3. Contingent Claims methods
• Does the credit union have at least 5 years of monthly deposit balance data with
which it can estimate the necessary parameters?
The five-year requirement is to ensure that parameter estimates used in these

sophisticated methods have a reasonable degree of statistical accuracy.

• Using this data, can the credit union demonstrate a contingent claims-based
method is appropriate for their institution?
This has to do with the degree of interest rate sensitivity of the credit union’s deposit
balances. The less the institution’s propensity to change deposit rates as market rates
change, and the more customer loyalty (less interest rate sensitivity of balances to
changes in market rates), the less the embedded options are a factor in valuing the
deposits. Contingent claims are therefore best applied in institutions that can
demonstrate a significant degree of interest rate sensitivity in the deposit rates and
balances.
• Market data is typically required as inputs for these models. Does the credit
union have access to a reliable and timely source of such information?
These more sophisticated methods of valuing NMDs will only yield accurate estimates
of the value of the deposits if the parameters estimates are based on accurate data.
Reliable sources for market data include commercial services such as Bloomberg, and
Reuters. OTS publishes the Treasury spot rate (zero-coupon) yield curve prevailing at
the end of each quarter in its publication Selected Asset and Liability Price Tables.
• Were the parameters updated at the time of the most recent valuation, using up-
to-date data?
Since many of the parameters used in these sophisticated methods may vary in
different economic and interest rate scenarios, it is important that each time the
valuation is performed the parameters used reflect the most recent and up-to-date
information possible. This includes both market data and data specific to the credit
union.
• Does the method used to model future deposit balances reflect expected growth
of the deposit balances?
As discussed in the Introduction and in chapter II, the NEV model is not consistent with
modeling growth in deposit balances.

REFERENCES
Alger, Guillermo, and Ingela Alger, “Liquid Assets in Banks: Theory and Practice”, unpublished
working paper, Boston College, 1999.
Black, Fisher, and Myron Scholes, “The Pricing of Options and Corporate Liabilities,” Journal
of Political Economy, Vol. 81, 1973, 637-659.
Board of Governors of the Federal Reserve. “Interagency Guidance on Sound Practices for
Managing Interest Rate Risk”. SR-96-13
Chen, Ren-Raw, and Louis Scott, “Pricing Interest Rate Options in a Two-Factor Cox-
Ingersoll-Ross Model of the Term Structure, Review of Financial Studies, Vol. 5, No. 4,
1992, pp. 613-636.
Duffy, Peter, “In the margins”, Credit Union Management, October 2000, pp. 18-22.
Elliot, Rober J., and P. Ekkehard Kopp, Mathematics of Financial Market. New York, NY.
Springer-Verlag New York, Inc.
Evans, David S., and Bernard Shull, “Economic Role of Credit Unions in Consumer Banking
Markets”, consulting report prepared for Boeing Employees’ Credit Union, National
Economic Research Associates, January 1998.
Fabozzi, Frank J. 2000. Fixed Income Analysis for the Chartered Financial Analyst Program.
New Hope, PA: Frank J. Fabozzi Associates.
Ferguson, Robert, and Raymar, Steven, “A Comparative Analysis of Several Popular Term
Structure Estimation Models”, Journal of Fixed Income, March 1998, pp.17-34.
FFIEC. “Supervisory Policy Statement on Investment Securities and End-User Derivatives

Activities” (63 Fed. Reg. 2019 at 20195), April 23, 1998.
Goosse, Philippe, Philippe Grégoire, and Isabelle Platten, “Interest rate risk in non-maturity
banking products: Implied options in savings accounts”, unpublished working paper,
l’Université Catholique de Louvain, 1999.
Hull, John C., Options, Futures, & Other Derivatives, 4th edition. Upper Saddle River, NJ:
Prentice Hall, 2000.
Hutchison, David, “Retail Bank Deposit Pricing: An Intertemporal Asset Pricing Approach”,
Journal of Money, Credit and Banking, Vol. 27, February 1995, pp. 217-231.
______________, and George Pennacchi, “Measuring Rents and Interest Rate Risk in
Imperfect Financial Markets: The Case of Retail Deposits”, Journal of Financial and
Quantitative Analysis, Vol. 31, September 1996, pp. 399-417.
Ide, David J., “The Maturation of Nonmaturity Deposit Modeling”, Bank Accounting and
Finance, Winter 1998-99, pp. 7-12.
Janosi, Tibor, Jarrow, Robert and Ferdinando Zullo, “An Empirical Analysis of the Jarrow-van
Deventer Model for Valuing Non-Maturity Deposits”, Journal of Derivatives, Vol. 7, Fall
1999, pp. 8-31.
Jarrow, Robert A., and Donald R. van Deventer, “The arbitrage-free valuation and hedging of
demand deposits and credit card loans”, Journal of Banking and Finance vol. 22, 1998,
pp. 249-272.
Kahn, Charles, George Pennacchi, and Ben Sopranzetti, “Bank Deposit Rate Clustering:
Theory and Empirical Evidence”, Journal of Finance vol. 54, December 1999, pp.
2185-2214.
Mays, Elizabeth, “The Estimation of the Duration of Nonmaturity Deposits”, in “Controlling and
Managing Interest-Rate Risk, Anthony G. Cornyn, Robert A. Klein, and Jess
Leiderman, eds., New York Institute of Finance, 1997, pp. 70-88.
McGuire, William J., “The Evolution of Interest-Rate Risk Measurement in Financial

Institutions”, in “Controlling and Managing Interest-Rate Risk, Anthony G. Cornyn,
Robert A. Klein, and Jess Leiderman, eds., New York Institute of Finance, 1997, pp.
56-69.
________________, and Skip J. Massengill, “Comprehensive asset/liability management:

More than just a model”, Bank Accounting and Finance, Winter 1996/1997, pp. 21-29.
________________, and T. Jamie Newell, “Modeling Core Deposits: The Commerce Bank
Methodology”, Chapter 8 in Interest Rate Models: Theory and Practice, Anthony G.
Cornyn and Elizabeth Mays, eds., Glenlake Publishing Company, 1997, pp. 109-146.
NCUA. “Investment Securities and End-User Derivatives: Interpretative Ruling and Policy
Statement” (63 Fed. Reg. 24097 at 24101) May 1, 1998.
O’Brien, James M., “Estimating the Value and Interest Rate Risk of Interest-Bearing
Transactions Deposits”, unpublished working paper, Board of Governors of the Federal
Reserve, November 2000.
_______________, Orphanides, Athanasios, and David Small, “Estimating the Interest Rate
Sensitivity of Liquid Retail Deposits”, Proceedings of a Conference on Bank Structure
and Competition, Chicago Federal Reserve Bank, May 1994, pp. 400-435.
Office of the Comptroller of the Currency, “Risk Modeling and Model Validation”, Bulletin 2000-
16.
Office of Thrift Supervision, “The OTS Net Portfolio Value Model”, 1994.
______________________, “The OTS Net Portfolio Value Model”, 2001.

Rebonato, Riccardo, Interest Rate Option Models, 2nd edition. New York: John Wiley & Sons,
1998.
Rigsbee, Stephen R., Sirri S. Ayaydin, and Charles Richard, “Implementing the Option-
Adjusted Spread Model for Global Balance Sheet Management”, in “Controlling and
Managing Interest-Rate Risk, Anthony G. Cornyn, Robert A. Klein, and Jess
Leiderman, eds., New York Institute of Finance, 1997, pp. 358-392.
Selvaggio, Robert, “Using the OAS Methodology to Value and Hedge Commercial Bank Retail
Demand Deposit Premiums”, Handbook of Asset/Liability Management, Frank Fabbozi
and Atsuo Knoishi, eds., 1996, pp. 363-373.
Trigeorgis, Lenos, Real Options: Managerial Flexibility and Strategy in Resource Allocation.
Cambridge, MA. The MIT Press.
Tripp, James D., Stephen K. Lacewell, and Peppi M. Kenny, “What if CUS responded faster to
market-rate changes?”, Credit Union Executive, November/December 1997, pp. 28-35.
U.S. Treasury, Credit Unions: A Report to the U.S. Congress, 1997.
Van Deventer, Donald R., and Kenji Imai, “Nonmaturity Deposits”, Chapter 15 in Financial Risk
Analytics: A Term Structure Approach for Banking, Insurance and Investment
Management, Irwin Professional Publishing, 1997.
de Vassal, Vladimir, “Consumer Deposit Behavior and the Related Interest-Rate Risk”, in
“Controlling and Managing Interest-Rate Risk, Anthony G. Cornyn, Robert A. Klein, and
Jess Leiderman, eds., New York Institute of Finance, 1997, pp. 393-407.

APPENDIX
Summary of the Methods
This appendix contains a more detailed description of the various methods discussed in the
main body of the report. It is designed to provide a more technical background for those
readers who want deeper understanding of the methods and their assumptions. Proofs and
detailed discussions are omitted; for those the interested reader is referred to the original
publications.
To make things simpler for the reader, all the equations from the original papers have been
converted to use a common list of variables and definitions. They are provided in detail in
Table 12.
Table 12. List of Variables and Definitions
Variable Definition Variable Definition
D Account Balance c Institution’s non-interest

expense rate
i Interest rate paid on deposits f Reserve requirement
r Market rate g Decay rate for balances
R Alternative cost of funds Y Aggregate Income
Z Spot (zero-coupon) rate S Option-adjusted spread
All the methods of valuing NMDs involve four important building blocks. Before the cash flows
associated with the NMDs can be valued, the method must have a way of estimating future
deposit rates, future market rates, and future deposit balances. The final building block is the
valuation equation, which brings together the other three components. Each of the methods
discussed in Chapter II will be presented here in terms of the three primary building blocks and
how they are brought together to value the overall deposits.
To simplify the comparison of the different methods, rather than present and discuss each
method separately, we have grouped together all of the equations by the four building blocks
described in the previous paragraph. Thus, all of the processes used to estimate future market
rates are presented in Table 13, all of the equations to estimate future deposit rates are in
DRAFT CONFIDENTIAL
n/e/r/a
Consulting Economists
Table 14, the formulas for relating future changes in deposit balances are presented inTable 5,
and the overall valuation equations in Table 6.
Table 12. Market Rate Processes

Description: This table summarizes the interest rate processes assumed in each of the methods
discussed, as well as the interest rates used for discounting.
Valuation Method Market Rate Process Yield Curve Used

Hutchison & Pennacchi Vasicek Treasury
Jarrow & van Deventer Heath-Jarrow-Morton Treasury
(Kamakura Corporation)
McGuire Performance Solutions Base case = implied forward FHLB Advances curve
rates.
Additional scenarios considered.
Option of Monte Carlo
simulations of interest rate
paths.
O’Brien Cox-Ingersoll-Ross Treasury
OTS (1994) Spot (zero coupon) rates from Secondary CD
current yield curve.
OTS (2001) Spot (zero coupon) rates from LIBOR
current yield curve.
Selvaggio OAS Eurodollar

Table 13. Deposit Rate Processes
Description: This table summarizes the deposit rate processes assumed in each of the methods
discussed herein.
Valuation Method Deposit Rate Process

Hutchison & Pennacchi i = a 0 + a1 rt −1 + a 2 it −1 + ε t
t
it = i0 + β 0t + β 1 ∑ rt − j + β 2 (rt − r0 )
Jarrow & van Deventer
(Kamakura Corporation) j =0
McGuire Performance Institution specific.

Solutions
O’Brien (
i t = i t − 1 + λ + I t + λ − (1 − I t ))(E t − i t −1 ) + e t
1 if E t − i t − 1 > 0
It = 
 0 otherwise
E t = br t − g
OTS (1994) it = it −1 + a1 (it −1 − it − 2 ) + a2 (rt − rt −1 ) + a3 (rt −1 − rt − 2 ) + Bt −1 + Bt − 2
φ (i − Et −1 ), if it −1 ≥ Et −1
Bt −1 =  t −1
δ (it −1 − Et −1 ), if it −1 ≤ Et −1
γ (i − Et − 2 ), if it − 2 ≥ Et − 2
Bt − 2 =  t − 2
κ (it − 2 − Et − 2 ), if it − 2 ≤ Et − 2
Et = a + brt
OTS (2001) it = it −1 + a1 (it −1 − it − 2 ) + a2 (rt − rt −1 ) + a3 (rt −1 − rt − 2 ) + Bt −1
φ (it −1 − Et −1 ), if it −1 ≥ Et −1
Bt −1 = 
δ (it −1 − Et −1 ), if it −1 ≤ Et −1
Et = a + brt
Selvaggio n.a.

Table 14. Deposit Balance Processes
Description: This table summarizes the deposit balance processes assumed in each of the methods
discussed herein.
Valuation Method Deposit Balance Process

Hutchison & Pennacchi Dt e − µt = k1 rt + k 2 i t + η t
where dη t = [a 3 + b33η t ]dt + σ s 3 dz 3
t −1
Jarrow & van Deventer α 0t +α1t (t +1) / 2 +α 2 ∑ rt − j +α 3 (rt −r0 )
(Kamakura Corporation) Dt = D0 e j =0

Solutions
O’Brien log(Dt ) = α 1 + α 2 (rt − it ) + α 3 log(Yt ) + α 3 log(Dt −1 ) + vt
 
OTS (1994)
i
 it

  ( q +α ( s − q ) / t )  1200
µ

− m*  e q + 1  ( )
Dt = Dt −1  t  ⋅ e +  
 rt   m *
+ 2 (  )
 
 
1
OTS (2001)
  i  12
Dt = Dt −1  a + b × arctan d + c × t  + e × rt 
 
  rt  
Selvaggio log(Dt ) = α1 log(Dt −1 ) + α 2 log(rt ) + α 3 log(Yt ) + α 4 (TimeTrend ) + α 5 (MonthDummies ) + vt

Table 15. Valuation Equations
Valuation Method Valuation Equation

T
Hutchison & Pennacchi
PV = ∫ P (t )((1 − f )rt − it − c )Dt dt
0
 1 − e b11t
where P (t ) = EXP 
( )
(rt − r∞ ) − tr∞ + σ 13 1 − e b11t 2 
2

( )
 b11 4b11 
where r∞ = − a1 b11 − qσ 1 b11 − 2 σ 1 b11
1 2 2
τ
Jarrow & van Deventer
PV = cD0 ∫ (1 c ) e µ 0 + µ1t −σ 1t / 2 ×
εt 2
(Kamakura Corporation) 0
{(1 − m)[µ 3t [ ]}
+ σ 13,t ] − k + π t (i0 − k ) + µ 2t + σ 12,t dt
Solutions
O’Brien (it −1 + ct −1 )Dt −1 − (1 − f )∆Dt
∞
PV = fD0 + ∑ E 0Q
t =1 (1 + Z t )t
OTS (1994) 360
cDt −1 + [(1 + it )Dt −1 − Dt ]
PV = ∑t =1 (1 + Z t )t
OTS (2001) 360
cD + [(1 + it )Dt −1 − Dt ]
PV = ∑ t −1
t =1 (1 + Z t )t
Selvaggio 1 360  (1 − f )rt − c 
DDP0 = ∑  D(rt −1 ) 
12 t =1  (1 + Z t + S σ / 12)t 

GLOSSARY OF TERMS
Measure of Effective Description

Maturity
Decay rate The decay rate is the proportion of an account that can be
expected to be withdrawn over the next year.
Deposit premium The difference between the face value of the deposit and the
present value of the economic rents attributable to the
deposit.
Discount rate The interest rate used to calculate the present value of a
future cash flow.
Duration The percentage change in deposit value for a small change in
interest rates.
Dynamic discounted cash To calculate the present value of a stream of cash flows over
flow different interest rate paths, taking into account how the cash
flows will vary as interest rates change. Usually accomplished
using Monte Carlo simulation, but also with multiple interest
rate scenarios.
Economic rent The net benefit to a deposit-taking institution from being able
to issue deposits paying less than comparable-risk
investments. Usually measured as the present value of the
spread between the market rate and the deposit rate,
adjusted for the non-interest service costs.
Effective maturity A final maturity chosen for a particular model.
Option-adjusted spread The amount that must be added to the benchmark discount
rate (usually the risk-free rate), so that the present value of
expected future cash flows calculated over many different
interest rate paths equals the observed market value.
Parameter The coefficient from a regression that relates the sensitivity
of, say, deposit balances or deposit rates to changes in other
variables. Alternatively, the coefficient in a valuation model
that translates changes in input variables into valuation
effects.
Retention rate The retention rate is the proportion of an account that can be
expected to still be in the institution one year later.
Static discounted cash flow To calculate the present value of a single deterministic
stream of cash flows using a single interest rate path, without
modeling what might happen to interest rates.

The Evaluation of Credit Union Non-Maturity Deposits

Uploaded by

Copyright:

Available Formats

The Evaluation of Credit Union Non-Maturity Deposits

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Evaluation of Credit Union Non-Maturity Deposits

Uploaded by

Copyright:

Available Formats

SEP.

THE EVALUATION OF CREDIT UNION NON-MATURITY DEPOSITS

National Credit Union Administration

David M. Ellis, Ph.D., Senior Consultant

TABLE OF CONTENTS .................................................................................................................. I

LIST OF FIGURES ....................................................................................................................... III

LIST OF TABLES ......................................................................................................................... III

I. INTRODUCTION AND SUMMARY OF RECOMMENDATIONS ............................................................ 1

II. THE DIFFERENT METHODS FOR VALUING NON-MATURITY DEPOSITS ........................................ 8

III. CREDIT UNIONS VS. BANKS ................................................................................................. 30

IV. THE EFFECTIVE MATURITY OF NON-MATURITY DEPOSITS ..................................................... 37

V. DATA AND ESTIMATION REQUIREMENTS OF DIFFERENT METHODS .......................................... 44

VI. AN ILLUSTRATIVE SIMPLE MODEL ........................................................................................ 52

VII. RECOMMENDATIONS .......................................................................................................... 62

VIII. IMPLICATIONS AND GUIDELINES FOR NCUA ....................................................................... 67

Glossary of Terms ............................................................................................................... 80

Credit Union Non-Maturity Deposits ii

We were asked to include the following tasks in this evaluation:

• Comprehensive review of the literature on valuing NMDs.

• Conceptual evaluation of alternative methods.

• Analysis of the material costs and benefits of alternative methods.

Our recommendations are based on four main criteria:

• Conceptual consistency with the theory of NMD valuation.

• Applicability to credit unions.

• Feasibility of implementation based on availability of data and ease of estimation and

3. Valuation methods based on an “alternative cost of funds” approach tend to overstate

Credit Union Non-Maturity Deposits 2

4. Some of the simpler methods of including NMDs in asset-liability management systems

Table 1. Recommended Durations for NMDs

Credit Union Non-Maturity Deposits 3

7. We make no recommendations about the effective maturities in relatively fully-specified

b. Model parameters (the constants in the equations) should be estimated with

Credit Union Non-Maturity Deposits 4

B. Background on the History of Non-Maturity Deposit Valuation

In addition to the lack of competitive pressures on deposit taking institutions to offer

Credit Union Non-Maturity Deposits 5

Credit Union Non-Maturity Deposits 6

Credit Union Non-Maturity Deposits 7

Credit Union Non-Maturity Deposits 8

Credit Union Non-Maturity Deposits 9

Credit Union Non-Maturity Deposits 10

A. Overview of Valuation Theory Applied to NMDs

Credit Union Non-Maturity Deposits 11

Thus, a fully-specified NMD model consists of the following:

• A theoretically sound method of discounting, either discounting expected cash

Credit Union Non-Maturity Deposits 12

B. Economic Rents as the Source of NMD Premia

Credit Union Non-Maturity Deposits 13

 Interest that could be received  Interest   Non − interest 

 Net Present Value 

Credit Union Non-Maturity Deposits 14

The net liability of the deposit can then be written

 Net Present Value 

C. Valuing Deposit Liabilities Based on Economic Rents

Vd = deposit liability value.

D0 = par value of deposit.

i = credit union deposit rate, assumed constant in this derivation.

Credit Union Non-Maturity Deposits 15

Credit Union Non-Maturity Deposits 16