AN INVESTIGATION OF CASH MANAGEMENT

PRACTICES AND THEIR EFFECTS ON

THE DEMAND FOR MONEY
Michael Dotsey

I. INTRODUCTION The paper is organized as follows. Section II

provides an overview of some of the more popular
During the mid-1970s standard regressions ex- methods of economizing on cash balances. Section
plaining the demand for money underwent a- well- III discusses a number of proxies that have been
documented shift. 1 This shift was largely attributed employed in attempts to capture the effects of cash
to the adoption of more sophisticated methods of management on the demand for demand deposits.
cash management practices by firms. Specifically, Section IV analyzes the relative ability of these
techniques were developed that allowed firms to per- proxies to help explain cash management effects,
form a given level of transactions while holding lower while Section V provides a summary and conclusion.
average money balances. Therefore, for given levels
of transactions income and interest rates the demand
for money was lower than that implied by any his- II. POPULAR METHODS OF CASH MANAGEMENT
torical relationships.
This section provides an overview of some popular
This article investigates the effects that a number
techniques used by firms in economizing on cash bal-
of variables related to cash management have on the
ances. These techniques are an outgrowth of chang-
demand for money. These variables are generally
ing technology, as well as a response to the higher
suggested by analyzing specific methods of cash man-
opportunity costs of holding transactions or cash
agement and developing measures that incorporate
balances in the 1970s. The desire to economize on
the intensity and sophistication of these methods. All
the holdings of demand deposits has always been
of the variables examined help explain the shift in
present. However, changing economic conditions
the demand for money. Most notably, the number of
alter the profitability of investing in new methods of
electronic funds transfers made over the Federal
managing transactions balances. For example, lower
Reserves wire system restores stability to the esti-
computer costs may make previously unprofitable
mated demand for money function.
procedures profitable and advances in computer tech-
Since the use of more sophisticated cash manage-
nology may make new methods in cash management
ment techniques is believed to have its main effect on
feasible. Firms that are attempting to minimize the
demand deposit balances, this study concentrates on
costs involved in carrying out transactions, costs that
demand deposits. The empirical work uses annual
involve the interest foregone on idle balances, will
data over the period 1920-1979. Annual data is used
respond to changes in their economic environment by
to avoid controversy over the use of lagged dependent
altering the levels and types of cash management
variables. The starting date represents the earliest
services. The degree to which cash management
date for which reliable data on all variables could be
technology is employed will be arrived at in a manner
obtained, while the terminal date was selected to
analogous to the choice of any other investment
avoid the problems introduced by NOW accounts.
project.
This relatively long sample highlights the significant
effects that more sophisticated cash management The major changes that have spurred the growth
methods have had on the demand for demand de- of more sophisticated and more widespread use of
cash management techniques in the 1970s have been
posits.
the improvement of computer technology, the lower-
ing of computer costs, and the rise in market interest
For some examples see Lieberman [16], Kimball [12],
1

Porter, Simpson, and Mauskopf [19], and Simpson and

rates. These changes have increased the benefit for
Porter [20]. reducing transactions balances while lowering the

FEDERAL RESERVE BANK OF RICHMOND 3

costs of doing so. As a result, there has been a pro- Consolidating Cash Balances
liferation of ways in which firms manage their cash
balances. Examining these methods will indicate the Concentration Accounts Concentration accounts
exact ways that a firm can reduce its demand for allow firms to pool the balances collected by or de-
money. posited in local banks. Local banks automatically
The various methods used in managing cash bal- transfer funds, either by wire or by a depository
ances can be divided into three basic types. One type transfer check to a central concentration bank. This
speeds up the collection of receivables, another allows process is advantageous for a number of reasons. It
firms to consolidate accounts, while a third helps allows the firm to consolidate its cash balances,
control disbursements. Most of the techniques are making it easier and less expensive to switch idle
linked with improved means of accounting, enabling funds into market instruments. It also reduces the
the firm to more efficiently monitor its cash position. amount of total cash balances that need to be held
Also, many of the techniques are used together, since it allows for some offsetting of local distur-
thereby providing a full range of complementary ways bances to transactions balances.
for economizing on transactions balances.
Depository Transfer Checks Depository transfer
Methods For Speeding Up Receivables checks, like preauthorized checks, are a signatureless
check. They are issued by the concentration bank
Lock Boxes Essentially a lock box is a centrally against one of the firms local collection banks based
located collection post office box selected to minimize on deposit information sent from the collection bank
the mailing time taken to receive payments from cus- to the concentration bank, usually over a data pro-
tomers. A firm will usually operate a number of cessing network. Specifically, the concentration bank
lock boxes in various areas of the country. Several
receives the deposit data and issues a check the same
times a day, the firms local bank will open the lock
day for collection. It then sends the information con-
box, sort out the checks and deposit the money in the
cerning the collected funds and their availability to
firms account. The bank will then send the invoices
the firm.
and a record of the deposit to the firm. Often photo-
copies of checks will be sent and, in many cases, the Wire Transfers A wire transfer is a transfer of
information will be processed on magnetic tapes that funds most often sent over either the Federal Re-
can directly interface with the firms accounting serves wire system or a bank wire system. In this
system. For a given availability schedule, the firm case, the local bank transfers funds from the cor-
will have a good idea of the amount of money clearing porations account to the concentration bank. The
into each account on any given day. On average, lock wire transfers advantage is that it allows for same
boxes can reduce mail float from one to four days. day use of funds, while its disadvantage is that it is
Preauthorized Checks A preauthorized check is a somewhat more costly than depository transfer
signatureless check used for accelerating the collec- checks. Therefore, wire transfers are predominantly
tion of fixed payments. The customer signs an used for larger transfers than are depository transfer
agreement with the corporation allowing the cor- checks. For example, at an interest rate of 6 percent,
poration or the corporations bank to write a check at it would require a one-day transfer of $36,000 to
specified dates for specified amounts on his account. cover the typical $6.00 wire transfer cost. Naturally,
The corporation, through the use of a computer file, as interest rates rise, the minimum profitable level of
sends the bank the necessary information for per- the transfer would fall.
forming this function. The bank then informs the There have been a number of innovations in the
firm by means of a computer tape of the deposit and use of wire transfers. Most notable is the ability of a
the availability of the funds. This process lowers the firms cash manager to initiate a wire transfer from a
uncertainty in income flows as well as reducing mail computer terminal that either interfaces with the
float. banks computer controlled wire system or with the
Preauthorized Debit A preauthorized debit has data base of a third party that is used by the bank.
the same effect as a preauthorized check. In the case Often this service is linked with other cash manage-
of a preauthorized debit, the customers account is ment services, such as programs that forecast a com-
automatically debited on a specified date and funds panys cash flows, and produces wire transfers that
are electronically wired from the customers bank to are less costly and that provide hard copy verification
the firms bank. of funds transferred.

4 ECONOMIC REVIEW, SEPTEMBER/OCTOBER 1984

Methods For Controlling Disbursements For example, in inventory models of money de-
mand, either stochastic or nonstochastic, some of the
Controlled Disbursement A firm may also exert
investment in cash management services can be
more control over its cash balances by being able to
viewed as lowering transactions costs. For instance,
better predict disbursements on a day-to-day basis.
a firm having a cash management system that allows
The firm can achieve this by using a bank that re-
it to perform investments in repurchase agreements
ceives only one shipment of checks from the Federal
from a computer terminal has invested in a procedure
Reserve each morning. The bank informs the firm that greatly reduces transactions costs. In stochastic
of the value of checks drawn on its account and the inventory models, many of the cash management
firm then knows, usually before noon, how much of services can be viewed as ways for reducing the
its balances are unnecessary.
variance of cash flows (see Porter and Mauskopf
Zero Balance Accounts This procedure is a special [18]). Therefore, some key elements of the demand
case of controlled disbursement that allows the firm for money, namely transactions costs and the variance
to maintain zero transactions balances at a number of cash flows, are not exogenous variables from the
of banks from which it writes checks. When the standpoint of individual deposit holders, but are vari-
value of checks presented against the firms account ables that can be influenced by the level of cash man-
is tabulated, the appropriate amount of funds are agement sophistication.
wired from a central account. This allows the firm This line of reasoning implies that firms are simul-
to greatly economize on the level of balances held at taneously choosing the level of investment in cash
each disbursing bank, and provides centralized data management services and their average deposit bal-
on transactions. ances. Since a number of the parameters that influ-
ence the demand for demand deposits are functions of
Summary of Cash Management Services the level of cash management, the level of cash man-
agement should appear in the demand for money
It is clear from the description of the methods used equation. 2
in managing cash balances that many of these pro- Failure to include a measure of the effects of cash
cedures will be simultaneously employed. For management in the demand for demand deposits will
instance, a firm is likely to have zero balance arrange- therefore result in a seriously misspecified regression.
ments with local banks that also provide lock box As a result, coefficient estimates will be biased and
services. Also, the firm will use both depository predictions from the regression will be inaccurate in
transfer checks and wire transfers to facilitate the periods when cash management practices are chang-
quick movements of funds. Crucial to the desire to ing. Further the regression will appear to be unstable
economize on transactions balances is the ability to (leading one to believe that the demand for money is
invest these funds in short-term market instruments unstable), when in fact the instability is totally due to
at relatively low costs. Otherwise there would be no an omission on the part of the econometrician.
reason to incur the costs involved in reducing the In this section a number of variables for capturing
average level of balances. cash management technology are examined. These
candidates are generally related to the actual methods
used in cash management and to the underlying costs
Ill. METHODS OF CAPTURING and benefits associated with investing in techniques
CASH MANAGEMENT EFFECTS IN that help economize on transactions balances.
MONEY DEMAND EQUATIONS
A Time Trend (T) The first and simplest way to
The preceding discussion described how various represent cash management innovations in a money.
cash management techniques are able to reduce the demand equation is by use of a time trend. This was
demand for money. Therefore, failure to incorporate initially employed by Lieberman [16]. The moti-
cash management effects in a demand deposit regres- vation behind this variable is that the adoption of new
sion will result in a misspecified equation. Since the technology will be fairly uniform and proceed at an
degree to which cash management procedures are exponential rate. This procedure explicitly, treats the
used is a choice Variable of the firm and is related to
the cost and benefits of investing in these procedures, 2
The result of the optimization process by which firms
this misspecification will have serious consequences choose the level of cash management services and average
demand deposit balance is a two-equation system that is
for any estimated equation. recursive. For more detail see Dotsey [9].

FEDERAL RESERVE BANK OF RICHMOND 5

process of changes in cash management practices as
exogenous. It therefore omits from consideration any
economic forces, such as changes in costs or returns,
that would be expected to alter the rate at which cash
management techniques are implemented. However, where r t is the long-term bond rate (Moody Aaa)
it serves as a useful benchmark for comparing the and the + sign indicates that only positive values of
effects that more sophisticated methods of incorpo- the expression in parentheses are used. The variable
rating the consequences of cash management have is somewhat sensitive to the value of n chosen, so
had on the demand for money. One practical problem variables using n = 3, 4, 5, 6 were constructed. All
in using a time trend is choosing the starting date gave similar results and only the values for n=4 are
for the trend. reported. A graph of the RATCHET is depicted in
Figure 1.
A Ratchet Variable (RATCHET) In general, a One can see that the formulation given by equa-
firm would adopt new methods of cash management tion 1 captures the ideas behind the ratchet variable.
if the expected benefits outweigh the costs. That is, For example let n=4. Then the current value of
investing in a new cash management system would RATCHET is equal to last periods value plus an
involve the same considerations as investing in any additional term. The additional term reflects the
other project. The motivation behind the use of a value of todays long-term interest rate relative to its
ratchet variable constructed from interest rates is to average over the latest four periods. If todays rate
capture some of the economic conditions that would is higher than this average, then RATCHET in-
lead to firms implementing more sophisticated cash creases indicating an increase in investment in cash
management techniques. management services. If todays rate is lower than
Since much of the costs of employing innovations the average, then RATCHET remains the same as it
in cash management are start-up costs (e.g., putting was last period. This implies no new investment in
in the necessary computer hardware and software), cash management technology, and that todays level
it follows that once a new cash management system is of technology is the same as last periods level.
in place it will remain in operation until it is replaced
Although the ratchet variable possesses some useful
by more advanced technology. For the investment to
features, it does have certain limitations. It only
be profitable, the interest rate savings incurred from
considers the potential benefits of new technology but
lower average money balances must be substantial
and expected to last for some time. One would
therefore expect that major innovations would occur
when long-term interest rates are high relative to Figure 1
their past history, and that these innovations would
continue to affect the demand for money once they THE RATCHET VARIABLE
are initially adopted. Long-term rather than short- 1920 - 1979
term interest rates are the relevant variable, because
they indicate that a movement in interest rates is ex-
pected to persist. One is also interested in the move-
ment of long-term rates with respect to its past, since
upward movements will spur new investment in cash
management due to the increased return obtained
from economizing on transactions balances.
The preceding discussion suggests that a nonde-
creasing variable based on long-term interest rates,
which increases (or ratchets up) when rates are rela-
tively high, would be helpful in explaining changes in
cash management practices and hence changes in the
demand for money. The specific formulation investi-
gated in this study is the one derived by Simpson and
Porter [20]. Specifically,

6 ECONOMIC REVIEW, SEPTEMBER/OCTOBER 1984

not its cost. Furthermore, the benefits are only Figure 2

potential, since one doesnt know how much econo-

REAL PRICE OF
mization occurs as a result of new technology. Also,
OFFICE COMPUTING EQUIPMENT
the variable does not consider depreciation. 1920 - 1979
The Price of Office Computing and Accounting
Equipment (P) The discussion in Section II makes
it clear that much of the use of cash management
techniques involve computers and accounting equip-
ment. Therefore the costs of this equipment will be
closely related to the costs of cash management. In
constructing a variable that captures these costs, it is
important that the variable take into account adjust-
ment in quality. For example, a new computer model
may cost slightly more than the one it replaces, but
it may be able to perform many more operations in
much less time. In terms of what the computer
actually does, the newer model is much less expensive
than the older model even though its price may be
somewhat higher. A true index of the computers
cost will take account of the change in quality. Such
an index is referred to as a hedonic price index.
invested overnight: To implement this type of ac-
As the cost of technology falls, more firms will tivity often requires the use of immediately available
adopt the technology thus reducing the demand for funds. Therefore, much of the transfer of money is
demand deposits. Therefore the price of office com- done over either the Federal Reserves wire system
puting and accounting equipment could help to ex- or over Bank wire.
plain shifts in the demand for money induced by cash For instance, a firm may use a number of lock
management. However, the price variable does have boxes, have a zero balance account with a disburse-
certain limitations. It does not account for tech- ment bank, and a consolidation account with another
nology already in place, nor does it reflect depreci- bank. On any given day, funds would be wired from
ation. Further it does not consider changes in the the lock box collecting banks to the bank maintaining
benefits that occur from the implementation of new the consolidation account and from the consolidation
cash management services. Therefore, it would be account to the zero balance account. Funds may also
natural to use this variable in conjunction with a be wired to another bank for the purpose of exe-
cuting a repurchase agreement if it can not be done
ratchet variable.
with the consolidating bank. In general there is good
For the years 1956-1979 data on the hedonic price
reason to believe that the number of electronic funds
of office computing and accounting equipment was transfers is largely determined by the degree of cash
obtained from McKee [17]. Although his procedures management practices. Because of this relationship,
are somewhat rough, they are the best available. For the number of electronic fund transfers is a logical
the time period 1920-1955, it is assumed that the real variable for helping explain the shift in the demand
cost of technology remained constant at its 1956 level. for money (for more detail see Dotsey [9]).
A graph of this variable is given in Figure 2. The value for the number of electronic fund trans-
fers used is restricted to funds transfers made over
The Number of Electronic Fund Transfers (EFT) the Federal Reserves wire transfer system and is
The motivation for this variable is largely attributed depicted in Figure 3. Since there are other wire
to Kimball [12]. The use of many of the cash transfer systems this value is not totally accurate.
management techniques discussed in Section II in- However, it is believed that the time series properties
volves the rapid movement of money so that it may of the measure is not much different than what would
be invested in short-term market instruments. In be observed if data on total wire transfers could be
many cases idle transactions balances may only be obtained.

FEDERAL RESERVE BANK OF RICHMOND 7

 Figure 3
eludes installment retail credit, noninstallment retail
N U M B E R O F credit, credit outstanding on bank credit cards, credit
ELECTRONIC FUNDS TRANSFERS owed to gasoline companies, and check credit. The
1920 - 1979 letter e refers to the disturbance term.
Consumption expenditures are used to represent
transactions income, while RD captures the desira-
bility of holding a demand deposit.3 RS and RCP are
used to capture the return earned on alternative assets
held by different classes of economic agents. The
real wage rate is a proxy for the value of time and is
therefore related to transactions costs, while PCR
attempts to net out the percent of transactions income
spent via credit.4

3
The use of Kleins rate involves some empirical issues
that make interpreting its effect difficult (see Carlson and
F r e w [ 6 ] ) . However, to the extent that one believes
that corporations earn a competitive rate on their de-
posits, omission of RD leads to specification bias. (For
more detail see Dotsey [9], especially footnotes 6 and 7.)
4
Since the emphasis of this article is to illustrate the
effects of cash management practices on the demand for
demand deposits, equation 2 is not discussed in detail.
For a full discussion see Dotsey [S].

IV. EMPIRICAL RESULTS

In order to appreciate the severity of the effect of Table I

cash management on the demand for demand de-
REGRESSION ANAYLSIS FOR DEMAND DEPOSITS
posits, a regression explaining demand deposit be-
havior is run over the period 1920-1965, a period in
which cash management innovations are believed to
be unimportant. (An examination of Figures l-3
indicates that the various proxies are fairly constant
over this time span.) This regression is then rerun
over the extended sample period (1920-1979), and
the results are compared. This is depicted in Table I.
The regression equation examined is based on an
inventory model of the demand for money used in
Dotsey [8], [9]. Specifically,

where the letters LN refer to the natural log of a

particular variable (i.e., LNX equals the log of X).
The letter D represents the level of real demand de-
posits, C represents the level of real consumption
expenditures, RD is the own rate of return on de-
mand deposits calculated using Kleins [14] method-
ology, RS is a weighted average of the interest rate
on passbook savings accounts and money market
mutual fund shares, RCP is the commercial paper
rate, W is the real wage rate, and PCR is the ratio of
credit to consumption where the level of credit in-

8 ECONOMIC REVIEW, SEPTEMBER/OCTOBER 1984

 The results of the regression run over the period period 1965-1979), instead of moving from point A
1920-1965, yield coefficients that are consistent with to point B along D, there is a movement from point
an inventory model of money demand. The error A to point C. Thus, excluding the incorporation of
term does not exhibit any serial correlation and one cash management implies that demand deposits will
can not reject the stability of the regression. The appear to be less sensitive to changes in consumption.
tests for stability used were a standard F-test, a test A similar argument would apply to the real wage
using the cusum of squares statistic developed by rate. With respect to interest rates, however, the
Brown, Durbin and Evans [5], and a test for sta- effect of omitting cash management could be ambigu-
bility using the varying parameters model of Cooley ous. This is because interest rates both rose and fell
and Prescott [7]. 5 The regression coefficients also over the period 1966-1979. For example, consider
converge fairly quickly to their full sample values, the commercial paper rate. Failure to explicitly
when the sample period is continually extended include the effects of cash management would imply a
from 1928 to 1965. This combination of evidence greater sensitivity of demand deposits to increases in
strongly implies that the specification in equation 2 the commercial paper rates when this rate was rising,
is a well-behaved representation of the demand for while just the opposite would occur when the rate
demand deposits over the period 1920-1965. was falling.
When the sample period is extended through 1979
this is no longer the case, and equation 2 is no longer Adding Proxies For Cash Management If a move-
an accurate model of the demand for demand de- ment toward more sophisticated cash management
posits. Most importantly, the error structure of the techniques is the sole or primary reason for a shift in,
regression changes. This is evident from the low the demand for demand deposits, then incorporating
value of the Durbin-Watson statistic, implying serial variables that accurately account for this movement
correlation in the errors. This means that the stan- should have a pronounced effect on estimated demand
dard errors of the regression coefficients are biased for demand deposit equations. Specifically, the errors
making it impossible to state whether the coefficients should be white noise and the coefficients should
in column 2 of Table I are significant. After cor- return to the values found in the regression run over
recting for serial correlation the wage variable the 1920-1965 period. Further, stability of the re-
becomes insignificant. Also, the presence of serial gression over the extended 1920-1979 sample should
correlation is often indicative of a missing variable
or variables. A good candidate, or candidates, for Figure 4
this missing variable would be variables that take
into consideration the effects of cash management. AN EXAMPLE OF HOW CASH
The reduction of the coefficients on consumption MANAGEMENT AFFECTS THE
and the real wage is consistent with the omission of MEASURED RESPONSE OF DEMAND
variables that represent a general decrease in trans- DEPOSITS TO CONSUMPTION
actions costs, or a lowering of the variance of cash
flows associated with a given level of business. Con-
sider the graph in Figure 4. The locus of points
labelled D, represents a relationship between real
consumption C, and real demand deposit balances
with all other variables (i.e., interest rates, PCR,
and W) held fixed. The locus D represents the
same relationship depicted for a more sophisticated
use of cash management. As shown, less real bal-
ances are held for any level of consumption, interest
rates, real wages, and the intensity of credit pur-
chases. Now as consumption rises (as it did over the

5
Descriptions of these test statistics are quite technical
and are therefore omitted. The interested reader can
find a more detailed discussion in Dotsey [8] or can read
the referenced articles. An excellent summary can also
be found in Boughton [4]. Demand Deposits

FEDERAL RESERVE BANK OF RICHMOND 9

not be rejected and there should be a significant of first order serial correlation of the residuals 6
increase in the predictive power of the equation. The Since RATCHET proxies for the potential benefits
various proxies described in Section III will be
examined with respect to all of the properties just 6
The discussion of EFT in Section III indicates that its
listed. effects should be examined within the context of a simul-
taneous system. Since the analysis indicates that this
First, examine the regressions in Table II. All of system is recursive, simultaneity bias will occur only if
the errors in the two equations are correlated. A two-
the proxies return the coefficients approximately to stage least squares estimation technique gave similar
the values estimated in column 1, Table I. However, results to those obtained using OLS, implying that simul-
taneity bias is not a problem. For a more detailed dis-
only in the case of EFT, can one reject the presence cussion see Dotsey [9].

Table II

REGRESSION ANALYSIS FOR DEMAND DEPOSITS (1920-l 979)

WHEN VARIOUS PROXIES FOR CASH MANAGEMENT ARE INCLUDED

I II III IV

The numbers in parentheses are t-statistics.

* indicates significance at the 5 percent level.

**indicates significance at the 1 percent level.

RHO is the coefficient for first order autocorrelation.

10 ECONOMIC REVIEW, SEPTEMBER/OCTOBER 1984

of cash management, while LNP attempts to capture Table III

costs in adopting new technology, it would be natural THE LOGARITHM OF ONE PERIOD AHEAD
to use both variables simultaneously. This was FORECAST ERRORS
attempted, but only LNP retained its significance,
perhaps because these variables only reflect general None T RATCHET LNP EFT
trends and are therefore only picking up an overall
tendency toward increasing cash management so- 1966 - .19 .04 - .15 .00 .05
phistication. Finally a regression including LNP, 1967 - .24 .04 - .13 .00 .11
RATCHET, and EFT was run with only EFT
retaining its significance. 1968 - .16 .16 .08 .18 .18
Second, all of the proxies decrease the instability 1969 - .15 .14 .16 .17 .23
of the regressions in the sense that the cusum of
1970 - .18 .15 .19 .17 .21
square statistic is lowered. However, only by using
EFT could a lack of stability be rejected using the 1971 - .14 .10 .06 .12 .17
Brown-Durbin-Evans test. Also, one could not reject
1972 - .25 .05 - .12 - .03 .07
stability of the regression with EFT under the pro-
cedure developed by Cooley and Prescott. However, 1973 - .17 .07 - .04 .10 .32
when the sample was divided in 1949, stability was
1974 - .26 - .02 - .08 .03 .20
rejected using a standard F-test. Given that EFT
only includes the number of wire transfers over the 1975 - .71 - .04 - .47 - .38 - .28
Federal Reserve wire system, and is therefore an
1976 - .94 - .62 - .72 - .53 - .26
imperfect measure of total wire transfers, the net
result of the stability tests is encouraging. 1977 - .63 - .48 - .70 - .36 .02
Third, an examination of one step ahead out of 1978 - .73 - .48 - .80 - .33 .05
sample forecast errors is depicted in Table III.
Again, all the proxies generally improve the forecasts, 1979 - .69 - .30 - .60 - .40 .41
with EFT performing the best. Using EFT resulted Avg.
in a reduction of the average absolute error of the Error - .38 - .12 - .24 - .07 .11
forecast by 52 percent and a reduction in the root
RMSE .29 .28 .35 .24 .19
mean square error by 34 percent.

V. SUMMARY AND CONCLUSION

An attempt is made to capture this process by the
This article, in a somewhat different empirical use of variables which are believed to be related to
setting than that underlying most conventional studies innovations in the management of transactions bal-
of money demand, presents confirmation of the recent ances. This process seems to be captured quite well
shift in the demand for money. The hypothesis that a by the variable EFT. The other proxies perform
shift has taken place in the function is supported by reasonably well in reducing the forecasting errors of
stability tests and the poor predictive performance of the demand deposit relationship, but were not in
the model in the mid- and late 1970s. The empirical general able to capture the entire movement in the
evidence combined with documentation on the in- function. The results are by and large encouraging
creased use of sophisticated cash management prac- enough to make future research into transactions
tices by firms makes changes in cash management technology and its relation to money management a
techniques a probable explanation for the shift in the potentially rewarding avenue in helping to explain the
historical demand deposit relationship. current behavior of the demand for money.

FEDERAL RESERVE BANK OF RICHMOND 11

 References

1. Barger, H. Outlay and Income in the United States 12. Kimball, R. C. Wire Transfer and the Demand
1921-1938. New York: National Bureau of Eco- for Money. New England Economic Review,
nomic Research, 1942. Federal Reserve Bank of Boston (March/April
1980), pp. 5-22.
2. Baumol, W. J. The Transactions Demand for
Cash-An Inventory-Theoretic Approach. Quar- 13. Klein, B. The Payment of Interest on Commercial
terly Journal of Economics 66 (1952), 645-56. Bank Deposits and the Price of Money: A Study
of the Demand for Money. Unpublished Ph.D.
3. Beehler, Paul J. Contemporary Cash Management. dissertation. University of Chicago, 1970.
New York: John Wiley and Sons, Inc., 1978.
14. . Competitive Interest Payments on
4. Boughton, J. M. Recent Instability of the Demand Bank Deposits and the Long-Run Demand for
for Money : An International Perspective. South- Money. American Economic Review 64 (1971),
ern Economic Journal 48 (1981), 579-97. 931-49.
5. Brown, R. L.; J. Durbin; and J. M. Evans. Tech- 15. Kuznets, S. National Income and its Composition.
niques for Testing the Constancy of Regression New York: National Bureau of Economic Re-
Relationships Over Time. Journal of the Royal search, 1941.
Statistical Society, Series B 37 (1975), 149-63.
16. Lieberman, D. The Transactions Demand for
6. Carlson, J. A., and J. R. Frew. Money Demand Money and Technological Change. Review of
Responsiveness to the Rate of Return on Money: Economics and Statistics 59 (1977), 307-17.
A Methodological Critique. Journal of Political
Economy 88 (1980), 598-607. 17. McKee, M. J. "Computer Prices in the National
Accounts: Are Our Economic Problems a Com-
7. Cooley, T. F., and E. C. Prescott. Systematic puter Error. Council of Economic Advisors, 1981.
(nonrandom) Variation Models Varying Parame- (Processed.)
ter Regression : A Theory and Some Applications.
Annals of Economic and Social Measurement 2 18. Porter, R. D., and E. Mauskopf. Cash Manage-
(1973), 463-73. ment and the Recent Shift in the Demand for
Demand Deposits. Board of Governors of the
8. Dotsey, Michael. The Effects of Cash Management Federal Reserve System, 1978. (Processed.)
Practices on the Demand for Demand Deposits.
Working Paper 83-2. Federal Reserve Bank of 19. ; T. D. Simpson; and E. Mauskopf.
Richmond, 1983. Financial Innovation and the Monetary Aggre-
gates. Brookings Papers on Economic Activity
9. The Effects of Cash Management (1: 1979)) 213-29.
Practices on the Demand for Demand Deposits.
1984. (Processed.) 20. Simpson, T. D., and R. D. Porter. Some Issues
Involving the Definition of and Interpretation of
10. ; S. Englander; and J. Partlan. Money the Monetary Aggregates in Controlling the Mone-
Market Mutual Funds and Monetary Control. tary Aggregates III. Federal Reserve Bank of
Quarterly Review, Federal Reserve Bank of New Boston Conference Series no. 23 (1980), pp. 161-
York 6 (1982), 9-17. 234.
11. Karni. E. The Value of Time and the Demand for 21. Tobin. J. The Interest Elasticity of Transactions
Money. Journal of Money, Credit and Banking 6 Demand for Cash. Review of Economics and Sta-
(1974), 45-64. tistics 38 (1953), 241-47.

12 ECONOMIC REVIEW, SEPTEMBER/OCTOBER 1984