«Evidence of Early Withdrawal in Time Deposit Portfolios JAMES H. GILKESON Assistant Professor, Department of Finance, University of Central Florida, ...»
Journal of Financial Services Research 15:2 103±122 (1999)
# 1999 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands.
Evidence of Early Withdrawal in Time
JAMES H. GILKESON
Assistant Professor, Department of Finance, University of Central Florida, Orlando
JOHN A. LIST
Assistant Professor, Department of Economics, University of Central Florida, Orlando
CRAIG K. RUFF
Vice President, Educational Products, Association for Investment Management and Research, Charlottesville, VA Abstract The embedded options found in some securities are known to have signi®cant impact on product pricing, secondary market valuation, and risk measurement and management. The option to withdraw commonly found in bank deposits is one of the least studied of these. We help to ®ll this gap by examining the level and interest rate sensitivity of early withdrawals of retail time deposits using panel data from the Thrift Financial Report. We ®nd that longer-maturity time deposit portfolios commonly experience early withdrawals at economically signi®cant levels. Further, we ®nd that depositors respond positively, with increased levels of early withdrawal, to the reinvestment incentive they face when new deposit rates rise. These ®ndings increase our understanding of consumer behavior with regard to ®nancial products and have signi®cant implications for the competitive pricing of deposit products and the management of bank interest rate risk.
Banking changed signi®cantly during the 1980s and 1990s, as ®nancial markets became increasingly volatile and competitionÐboth interbank and between commercial banks, investment banks, insurance companies, mutual funds, and other ®nancial service providersÐgrew.1 After years of losing depositors to mutual funds and other investment vehicles and losing borrowers to the increasingly accessible commercial paper and corporate bond markets and to mortgage brokers and ®nance companies, bankers have come to recognize that their customers are not captive.2 In response, they have become increasingly sales focused, paying much attention to offering competitive products and pricing. Further, increased market volatility has led bankers and their regulators to focus more effort on understanding the behavior of their assets, liabilities, and off-balance sheet activities (i.e., the behavior of their customers) under a wide variety of potential economic scenarios and competitive situations.
Despite the downward trend, deposits remain the most important source of bank funding, accounting for 68% of domestically chartered commercial bank assets in May
1998. These deposits are cheap relative to other funding sources and allow banks to 104 JAMES H. GILKESON ET AL.
maintain decent interest rate spreads. Forevidence of the value of deposit funding, note that nonbank banks (®nance companies, mortgage banks, and the like) offer all manner of loan products. In addition, money market funds and other investment vehicles also compete directly with the liability products of banks. The unique bene®t extended to chartered banks and thrifts is the ability to offer FDIC-insured, retail (i.e., small) deposits.
Yet, despite complaints from some bankers about overly burdensome regulations and supervisory practices, there has been no rush to renounce charters. Appealing to banks' revealed preferences, we argue that the retail deposit franchise must be quite valuable.
The continuing dominance of deposits as a source of bank funding provides motivation for furthering our understanding of the behavior of depositors. After all, during the Regulation Q era, when deposit rate ceilings were in place, retail depositors' decisions were motivated primarily by branch convenience and noninterest premiums (those infamous toasters and electric blankets). However, since the early 1980s, insured bank deposits have been free from price controls. Of course, the growth of mutual funds, including money market and Treasury bond funds, which generally mimic the default-free character provided deposits by FDIC insurance, has brought additional competitive pressures. The growth of these funds and the relatively ¯at volume of deposits suggests a signi®cant migration of consumer investment dollars away from bank deposits. Time depositors in particular appear to have become increasingly interest rate sensitive in the 1980s and 1990s (see, e.g., Davis and Korobow (1987)).
Retail deposits are different from strictly ®xed-income securities because of the withdrawal option embedded or bundled with each account. Just as most residential mortgages allow the borrower to prepay the loan at will and without cost, retail (insured) deposits typically allow the depositor to withdraw funds at will. Demand deposits, such as savings, checking, and MMDA accounts, have no stated maturity but allow costless withdrawal. Time deposits allow early withdrawal prior to stated maturity but assess an early withdrawal penalty typically equal to some percentage of face value (although it may be stated as some number of months of interest).
While mortgage prepayment behavior has been studied in great detail, much less work has been done on the withdrawal behavior of bank depositors.3 This primarily is a result of the lack of a developed secondary market for retail deposit portfolios. Retail deposit portfolios typically sell only as part of the sale of a bank branch. In such cases, it is not possible to allocate the total sales price across the varied deposit portfolios and physical assets that were bundled together as the branch. Further, the bank regulatory agencies historically have not gathered data on depositor withdrawal behavior. In contrast, the development of an active secondary market for mortgage portfolios and mortgage-backed securities has spurred much interest in understanding the values and risks associated with mortgages, particularly prepayment risk. At the same time, the development of these secondary mortgage markets led to a large volume of market price and prepayment data, making such research feasible.
Although many mortgage prepayments are motivated by ®nancial incentivesÐthat is, the ability of borrowers to re®nance at a lower rate and reduce their monthly paymentsÐ there are many idiosyncratic or borrower speci®c motivations. These include the desire to move locally into a different home or change employment to a distant location. Similarly, while some early time deposit withdrawals may be motivated by ®nancial reinvestment
EARLY WITHDRAW IN TIME DEPOSITSAL incentives, it is likely that many are motivated by depositor liquidity needs. A time depositor who faces a sudden need for cash must choose between borrowing to meet that need or paying a penalty to withdraw some or all of his or her funds prior to the stated maturity. It often may be less expensive to pay the penalty than to pay the difference between the higher borrowing rate (e.g., a credit card rate) and the rate being earned on the time deposit. Our results show that early time deposit withdrawals occur at economically signi®cant levels and are motivated, in part, by the level of reinvestment incentive.
However, they also show that, on average, time depositors that withdraw funds early pay a substantial net penalty (i.e., they withdraw funds despite a negative reinvestment incentive). This suggests that liquidity needs play a substantial role in early withdrawal decisions.
This paper provides evidence on the early withdrawal behavior of time depositors and the sensitivity of early withdrawals to changes in interest rates. In this effort, we study data that was recently collected by the Of®ce of Thrift Supervision (OTS). In the ®rst section of this paper, we review the existing literature regarding the pricing and risks of bank deposits and the behavior of depositors. In the second and third sections, we describe the theory we test and summarize the data we use. We describe our speci®c empirical methodology in the fourth section and the results of our tests in the ®fth. We end with a brief discussion of the implications of our work and suggestions for further research.
1. Bank deposits in the academic literature
The consumer deposit market in the United States is characterized by many small depositors, each of whom chooses to deposit (invest) funds in one of a handful of competitive banks. Each of these banks determines its deposit rates endogenously after considering current market interest rates, the current rates and likely rate responses of its competitors, the likely behavior of its current and potential depositors, and its current needs for various types of funding (e.g., loan demand). Deposit rates therefore are administered rates rather than market rates, like those on U.S. Treasury securities.
Previous studies involving consumer deposits have provided a number of insights about the pricing and risks of bank deposits and the behavior of depositors. These ®ndings are best summarized as a series of stylized ``facts'' that are supported by multiple studies.
1. Bank deposit rates generally follow market interest rates. A number of studies have shown that movements in market rates, in particular U.S. Treasury rates, explain much of the movement in bank deposit rates over time. Support for this idea is found in Cooperman, Lee, and Wolfe (1992); Diebold and Sharpe (1990); Gilkeson and Porter (1998); Mahoney, et al. (1987); Wenninger (1986); and others. These studies also show that some deposit rates, including those on time deposits and MMDAs, more closely follow market rates than others, such as those on checking, NOW, and savings accounts. In addition, evidence ®nds that smaller banks tend to follow the rate decisions of larger banks in their local market (see, e.g., Cooperman, Lee, and Lesage, 1990, and Hanweck and Rhoades, 1984).
106 JAMES H. GILKESON ET AL.
2. Bank deposit rates are rigid. A number of studies have shown that changes in bank deposit rates lag behind changes in market interest rates. Thus, while deposit rates eventually follow market ratesÐat least to some degree, depending on the type of depositÐit may take a number of weeks or months before deposit rates fully re¯ect a particular change in market rates. Studies reporting this ®nding include Diebold and Sharpe (1990), Gilkeson and Porter (1998), Hannan (1994), Hannan and Berger (1991), Mahoney et al. (1987), Neumark and Sharpe (1992), and Passmore and Sparks (1993).
3. Bank deposit rates are upward sticky. Not only do changes in bank deposit rates lag behind changes in market rates, they do so asymmetrically. Speci®cally, deposit rates are slower to increase when market rates increase than to decrease when market rates decrease. This effect is more pronounced in more concentrated markets (i.e., those with the least competition among banks). Studies reporting this ®nding include Berger and Hannan (1989), Cooperman, Lee and Lesage (1991), Gilkeson and Porter (1998), Hannan (1994), Hannan and Berger (1992), and Neumark and Sharpe (1992).
These empirical observations suggest that depositor behavior allows banks to behave somewhat monopolistically (or at least oligopolistically) when setting their deposit rates.
Depositors are willing to accept lower than market rates for a number of potential reasons.
Among the most popular explanations offered are depositor switching costs, the value of noninterest services bundled with deposits, and the impact of disintermediation over the long term. Flannery (1982), Neumark and Sharpe (1992), and others suggest that deposit relationships, particularly transactions accounts, are time consuming to change, causing depositors to be somewhat indifferent to changes in the spread between deposit and market rates. For example, changing a checking account can involve visits to two banks, awaiting new checks, changing automatic deposit and electronic payment information, and other efforts.
Davis and Korobow (1987), Heffernan (1992), Zephirin (1994), and others note that banks provide valuable services to depositors, such as free checking services or convenient branch locations, which are in addition to the interest paid. Because the level of services can vary signi®cantly among banks in the same market, comparing deposit rates may not accurately compare the total returns earned.
Considering only depositor behavior, Gilkeson and Porter (1998) suggest that migration of the most interest rate sensitive depositors from bank deposits into money market funds and other market-priced products may have left banks with self-selected investors who are by nature less inclined to react to market-based rates than others. Taking a similar focus on depositor behavior, Jackson and Aber (1992) ®nd that banks can attract deposits by frequently changing their deposit rates, even if such changes do not accurately re¯ect changes in market rates. In addition, Athanassakos and Waschik (1997) ®nd that, while demand for long-term deposits depends most on the spread between an institution's own rates and those of its competitors, it is also sensitive to the bank's corporate identity, including its advertising expenditure.
A few theoretical papers developed models of bank deposit that produce the characteristics observed empirically; that is, rates that are somewhat and asymmetrically rigid with respect to market interest rates or balances that respond slowly to changes in