Full text of Economic Quarterly (Federal Reserve Bank of Richmond) : May-June 1985, Vol. 71

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

ARE NOW’S BEING USED AS
SAVINGS ACCOUNTS?
Timothy Q. Cook and Timothy D. Rowe

Recent years have witnessed a host of financial
innovations, one of the most important of which has
been the introduction of interest-bearing checking
accounts. Currently, there are two types of interestchecking accounts available: NOW accounts, which
have a 5¼ percent interest rate ceiling, and Super
NOW accounts, which are ceiling-free. (A more
complete description of the characteristics of these
two accounts is found in the Box.) In this article
“NOW accounts” refers to all interest-bearing checking accounts, including Super NOWs, unless otherwise noted.
Some observers have argued that consumers are
using interest-bearing checking accounts for both
savings and transactions purposes, causing a blurring
of the formerly sharp distinction between “transactions accounts” and “savings accounts.“1 The likelihood of consumers using NOWs as savings accounts
depends primarily on two factors : yield and liquidity.
The rates paid on NOWs have been (and should remain) substantially lower than the rates paid on alternative savings vehicles (see Box), giving consumers
an incentive to keep their savings in these other vehicles. Savings balances held in NOW accounts, however, are more easily accessed than balances held in
alternative savings vehicles. (Although, as noted in
the Box, there are savings alternatives that are very
liquid.) This greater liquidity might induce some
consumers to use NOWs as savings accounts even
though they pay a lower yield.” In general, this decision depends on each consumer’s preference for yield
versus liquidity and on the sacrifice in yield and the
gain in liquidity from using a NOW instead of an
1

See Hafer (1984) and Hetzel (1984).

alternative savings vehicle, To help determine if a
substantial number of consumers are, in practice,
using NOWs as savings accounts, this article presents
some direct evidence on how consumers are using
NOWs.
Whether consumers are using NOW accounts as
savings accounts has important implications for
monetary policy. Under current procedures the
Federal Reserve sets annual target ranges for three
different monetary aggregates-M1, M2, and M3.
M1, which is intended to be a measure of the public’s
transactions balances, includes currency, travelers
checks, balances in demand deposit accounts, and
balances in NOWs. NOWs have been a significant
portion of M1 since 1981 and comprised one-fourth
of M1 as of the end of 1984. M2 and M3, the broader
monetary aggregates, include M1 plus balances in
various categories of savings vehicles. Historically,
aggregate balances in these savings vehicles have
grown more rapidly than aggregate balances in
transactions accounts. For this reason the Federal
Reserve has generally set the annual target ranges
for M2 and M3 two to three percentage points higher
than the target range for M1. The use of NOWs as
both transactions and savings accounts could substantially alter the relationship between M1 and economic activity. Since balances in savings vehicles
grow more rapidly than transactions balances, balances in NOW accounts used for both savings and
transactions would grow at a faster rate than balances
in transactions accounts.3 As a result, the growth
rate of M1 consistent with a given growth rate of
economic activity would be higher than it was before
the introduction of NOWs. In other words, the use
of NOW accounts as savings accounts would raise

2

Also, some consumers might consolidate their transactions and savings balances in NOWs because of minimum balance requirements. For example, consider the
case of an individual with insufficient savings balances to
meet the minimum balance requirement on a money
market deposit account, which is currently $1,000 but
was $2,500 before January 1, 1985. This person, still
saving in a passbook, might find it worthwhile to consolidate his savings and transactions balances in a NOW
since the interest foregone on his savings balances would
be negligible.

3

Theoretically, savings balances should grow either more
rapidly than or at the same rate as transactions balances.
If savings balances grow at the same rate as transactions
halances, it would not matter if consumers are saving in
NOWs because saving in NOWs would not change the
long-run relationship between the growth rates of M1
and GNP. Historically, however, aggregate balances in
savings vehicles have grown more rapidly than aggregate
balances in transactions accounts.

FEDERAL RESERVE BANK OF RICHMOND

3

BOX

CHARACTERISTICS OF INTEREST-BEARING CHECKING ACCOUNTS

4

Interest-bearing checking accounts are available in two forms: NOW accounts and Super
NOW accounts. These accounts can only be
held by individuals and nonprofit organizations.
NOW accounts were authorized nationwide as
of the beginning of 1981 by the Depository
Institutions Deregulation and Monetary Control Act of 1980; previously they were available
only in New England. The interest rate on
NOW accounts is limited by regulation to 5¼
percent and almost all financial institutions pay
that rate. Super NOW accounts, which were
authorized as of January 5, 1983, have no interest rate ceiling and the rates paid vary across
financial institutions. Until 1985, Super NOWs
were required by regulation to have a minimum
balance of $2,500. At present the legal minimum is $1,000 although many institutions impose higher minimums. Similarly, although
there is no legal minimum on NOW accounts,
most institutions require minimum balances; a
common requirement is $500.
As mentioned in the text, consumers consider
two factors in deciding whether to use interestbearing checking accounts as savings accounts:
yield and liquidity. The savings vehicle which
is closest to a checking account in terms of
liquidity is the money market deposit account
(MMDA), which pays an unregulated rate
and is very liquid but has limited transactions
capabilities.* As is shown in the accompanying
chart, the rates paid on NOWs and Super
NOWs have been well below the yields available on MMDAs. Most interest-checking accounts are regular NOWs, whose 5¼ percent
ceiling rate was 3.6 percentage points less than
the average MMDA rate in 1984. The rates on
ceiling-free Super NOWs have also been well
below those paid on MMDAs-averaging 1.4

percentage points less in 1984. Furthermore,
it is likely that Super NOW rates will remain
lower because a Super NOW is a costlier
source of funds for a bank than is an MMDA.
Since banks are required to hold 12 percent of
Super NOW deposits as non-interest-bearing
reserves, they would be expected to offer on
Super NOW deposits no more than 88 percent
of the rate they pay on MMDAs. In fact, the
rate on Super NOWs in 1984 averaged only
84.1 percent of the rate on MMDAs, which is
even lower than suggested by this rule-ofthumb. This even lower average rate on Super
NOWs probably reflects the fact that banks
have to recover the additional costs of providing
checking services on Super NOW accounts.**

* MMDAs are allowed three checks and three telephone transfers per month, as well as an unlimited
number of withdrawals in person or through automatic teller machines.

** Corcoran and Wachtenheim (1984) provide a
detailed analysis of the relationship between the
yields a bank would be willing to offer on MMDAs
and Super NOWs.

RATES ON NOWs,
SUPER NOWs, AND MMDAs

Source: Federal Reserve Statistical Release, H.6

ECONOMIC REVIEW, MAY/JUNE 1985

the annual target for the growth rate of M1 compatible with any given growth rate of nominal GNP.
One approach to determining whether consumers
are using NOWs as savings accounts is to look at
the relationship between the growth rates of M1 and
GNP. If consumers are using NOWs as savings
accounts, then the growth rate of M1 should be
permanently higher relative to the growth rate of
GNP since the introduction of nationwide NOWs in
1981. Unfortunately, two data problems make it
difficult to evaluate whether this has happened.
First, only a short period has elapsed since the introduction of NOWs nationwide. Second, the relationship between M1 growth and GNP growth was
temporarily distorted from 1981 through 1983 by
consumers switching from regular checking accounts
into NOW accounts.4 As consumers switched into
NOWs from regular checking accounts, they had to
transfer funds from savings in order to meet the
higher minimum balance requirements on NOWs
and still have the same amount of funds available for
transactions. These one-time shifts of funds from
savings into NOWs temporarily boosted the growth
rate of M1 relative to GNP growth but did not
reflect the use of NOWs as savings accounts.5
Rather than looking at M1 and GNP growth rates,
this study evaluates two types of direct evidence on
how consumers are using NOWs. First, it examines
survey responses by consumers on their holdings of
different accounts. If consumers are using NOWs
as both transactions and savings accounts these survey responses should indicate that consumers have
consolidated their transactions and savings balances
in NOW accounts. Second, it looks at a number of
characteristics of NOWs, including (1) average balances, (2) transactions activity, and (3) seasonal
behavior. If consumers are using NOWs as savings
accounts, then these characteristics should partially
resemble the characteristics of other savings accounts,
such as money market deposit accounts and money
market fund accounts. Alternatively, if consumers
are using NOWs solely as transactions accounts,
then these characteristics should resemble those of
consumer demand deposit accounts. Overall these

two types of direct evidence provide very little support for the view that consumers are using NOWs
as savings accounts.

EVIDENCE ON CONSOLlDATlON OF
TRANSACTIONS AND SAVINGS ACCOUNTS

Perhaps the most direct way to determine if consumers are using NOWs for savings purposes is to
see whether they have combined their regular checking accounts and savings accounts into NOWs.
Evidence on the account combinations of consumers
is available from the 1983 Survey of Consumer Finances, which collected detailed financial data on
3,824 randomly selected households nationwide.6 Any
consumer that had consolidated savings and transactions balances in a NOW would have shown up in
the survey as having a NOW and no “savings account.” Savings account is here defined to be either a
regular savings account, a money market deposit account (MMDA), or a money market fund account
(MMF).
While it is not necessarily true that all consumers
without savings accounts are using their checking
accounts for savings purposes, the percentage of
survey respondents with NOWs but without savings
accounts can be used as an initial estimate of the
percentage of consumers that are using NOWs for
savings purposes. To make this estimate, the survey
results were classified into three groups on the basis
of whether a household’s main checking account was
(1) a regular checking account, (2) a NOW, or
(3) no checking account, and then each category was
divided into households with and without savings
accounts. The number of households in each group
is reported in Table I. Seventy-nine percent of the
households with NOWs also had savings accounts.
Furthermore, the 21 percent of NOW holders without separate savings accounts was only 2.4 percent
of the sample.7 Therefore, even if one assumes that
all NOW holders without savings accounts use their
NOWs for savings purposes, the effect on aggregate
M1 growth would be small.
6

4

In addition, the sharp decline in interest rates in 1982
raised M1 growth relative to GNP growth in late 1982
and early 1983. Radecki and Wenninger (1983) and Judd
(1983) provide discussions of this period.
5

To account for the effects of these flows on M1 growth
in 1981, the Federal Reserve “shift-adjusted” M1, thereby
implicitly raising the M1 target range. T h e M 1 t a r g e t
range for the current year reflects the view that the relationship between M1 and GNP has returned to a more
normal and predictable pattern.

Further detail on the survey is found in the Appendix.

7

To check if the results were biased because the survey
was stated in terms of households rather than individuals,
subsamples of single person households and one-or-twoperson households were examined. The percentages in
the subsamples were very similar to those for the whole
sample, leading to the conclusion that there was no bias.
For example, the number of NOW holders without
s a v i n g s a c c o u n t s w a s 2 . 8 p e r c e n t in t h e s i n g l e p e r s o n
household subsample and 3.0 percent in the one-or-twoperson household subsample.

FEDERAL RESERVE BANK OF RICHMOND

5

Table I

Table II

ACCOUNT COMBINATIONS HELD BY HOUSEHOLDS

MEDIAN BALANCE IN MAIN CHECKING ACCOUNT
BY TYPE OF ACCOUNT

Households
with a
Savings
Account 1

Main Checking
Account

Households
without a
Savings
Account

Main Checking
Account

Households
with a
Savings
Account 1

Households
without a
Savings
Account

Regular checking

$ 500

$ 300

NOW

$ 900

$1250

No Checking Account
Number
Percent of category
Percent of total2

322

485

39.8

60.1

8.6

12.9

Regular Checking
Number

1796

712

Percent of category

71.6

28.4

Percent of total

47.9

19.0

343

91

79.0

21.0

9.1

2.4

1
Savings accounts include regular savings accounts, MMDAs, and
MMF s .

source:
Board
1983.”

of

Governors,

“Survey of Consumer Finances,

NOW
Number
Percent of category
Percent of total

1
Savings accounts include regular savings accounts, MMDAs, and
MMFS.

2
Total equals 3749. The table omits 75 households from the full
sample: 44 that did not answer relevant questions and 31 whose
main checking account was a type not included in M1.

source:
Board
1983.”

of

Governors,

savings purposes. In sum, since only 2.4 percent of
the households in the survey fell into the category of
NOW holders without savings accounts and since
only some of those households appear to be using
their NOWs for savings purposes, it seems highly
unlikely that a significant percentage of households
are using NOWs for savings purposes.

“Survey of Consumer Finances,

EVIDENCE ON CHARACTERISTICS OF
NOW ACCOUNTS

A second way to evaluate whether consumers are
using NOWs as savings accounts is to compare the
characteristics of NOWs with those of alternative
transactions and savings accounts. Three characteristics for which data are available to make this comparison are average account size, transactions activity, and seasonal behavior.

Two other survey results cast doubt on the validity
of the assumption that all NOW holders who do not
have savings accounts use their NOWs for savings
purposes. First, as shown in Table I, a substantial
percentage of regular checking account holders held
no funds in savings accounts. This indicates that
many respondents either did not save at all or held
all of their savings in financial assets not included in
this study’s definition of savings account. Second,
one would expect consumers who had consolidated
savings and transactions balances in NOWs to have
significantly higher average balances than consumers
who were using NOWs purely for transactions purposes. The survey, however, provided only weak
evidence of such an effect. As Table II shows, the
sample median of the balances in NOWs was only
$350 higher for households with no savings accounts
than households with some type of savings account.8
This difference is fairly small, and indicates that only
a small proportion of the households with NOWs but
without savings accounts are using their NOWs for

Chart 1 shows the size of the average balances held
in Super NOW accounts, regular NOW accounts,
and regular consumer checking accounts.9 (Since
average balance data are available for both Super
NOWs and regular NOWs, this section discusses
these accounts separately.) The average balance in
Super NOWs is much higher than the average balance in regular NOWs, which is higher than the
average balance in personal demand deposits (DDs).
These differences might be used to argue that
interest-checking accounts include a substantial
amount of savings balances. There are a number of

8

9

The survey estimates of account sizes appear to be
biased downward, probably by the tendency of respondents to under-report dollar amounts. More reliable estimates of account sizes are presented below.

6

Average Account Size

Since the distribution of account balances is skewed to
the right, the median is more representative of the typical
balance than is the mean. Unfortunately, reliable estimates of median balances are not available.

ECONOMIC REVIEW, MAY/JUNE 1985

Figure 1
Chart 1

THE DECISION TO SWITCH
FROM A REGULAR CHECKING
ACCOUNT TO A NOW

AVERAGE BALANCES IN
INTEREST-CHECKING ACCOUNTS
AND REGULAR CHECKING ACCOUNTS

Sources. Board of Governors, Banking Section; and
Functional Cost Analysis.

reasons, however, why interest-bearing checking accounts would be expected to have higher average
balances even if they were solely composed of transactions balances. The most important reason is that,
due to the high minimum balance requirements on
interest-checking accounts, only persons with large
average balances find it worthwhile to use them. The
cost of switching to a NOW from a regular checking
account is the interest foregone on savings-type deposits used to satisfy the higher minimum balance
requirement, while the benefit is the higher yield
earned on transactions balances. Since the cost of
switching is unrelated to account size and the benefit
increases with account size, only persons with high
average transactions balances have the incentive to
switch from a regular checking account to a NOW.
This result is summarized in Figure 1. Only persons
with regular checking account balances greater than
B* profit by switching to NOWs. For this reason,

one would expect the average balance to be higher
in NOWs than DDs. Similarly, since Super NOWs
have higher minimum balance requirements than
regular NOWs (see Box), only holders of very large
balances profit by switching to Super NOWs.1 0
The behavior of average account sizes following
the introduction of NOW and Super NOW accounts
illustrates the importance of segregation by account
size as a determinant of average account size. The
average balance in personal demand deposits fell by
21 percent in real terms (from $1,489 to $1,182 in
1984 dollars) following the introduction of nationwide NOWs, indicating that much of the higher
average balance in regular NOWs was caused by
persons with large transactions balances switching
from DDs to NOWs. Similarly, the average balance
in regular NOWs fell by 18 percent in real terms
(from $5,701 to $4,662 in 1984 dollars) following
the introduction of Super NOWs in January 1983,
indicating that much of the higher average balance in
Super NOWs is a result of regular NOW holders
with large balances switching to Super NOWs.
In addition to causing this segregation by account
size, the characteristics of interest-bearing checking
accounts cause the same individual to hold a higher
average balance in an interest-checking account than
he would in a regular checking account. First, the
higher minimum balance requirement on a NOW
forces the individual to maintain a higher average
10

For a more elaborate treatment of this effect, see
Corcoran and Wachtenheim (1984).

FEDERAL RESERVE BANK OF RICHMOND

7

balance in order to have the same amount of funds
available for transactions. Second, because NOWs
pay explicit interest, the marginal opportunity cost
of holding NOW deposits is much lower than that
of holding demand deposits. Hence, in general one
would expect an individual to hold more transactions
balances (above the minimum requirement) in a
NOW than in a demand deposit. Similarly, he would
hold an even higher balance in a Super NOW than
he would in a regular NOW, since Super NOWs
have higher required minimums and pay higher rates
than regular NOWs.

slightly since the first quarter of 1983.12 In general,
except during periods of regulatory change, the average balances in regular NOWs and Super NOWs
have been stable or decreasing and the average balance in personal demand deposits has been very
stable. This similarity in the movement in the average balances in interest-checking accounts and personal demand deposits suggests that interest-checking
accounts are not being used extensively for savings
purposes.

While these factors explain why interest-bearing
checking accounts would be expected to have a higher
average account size than consumer demand deposits,
they do not rule out the possibility that some of the
higher average balance in interest-checking accounts
results from some consumers using them as savings
accounts. If consumers are holding savings in
interest-checking accounts, however, the average
interest-checking balance should be increasing faster
over time than the average personal demand deposit
balance. The reason is that, as discussed above,
aggregate balances in savings vehicles have grown
more rapidly than aggregate balances in transactions
accounts. This difference in the behavior of aggregate balances probably reflects the tendency for individual savings balances to grow at a faster rate than
individual transactions balances.11 Therefore, if consumers are using NOWs partially as savings accounts, one would expect to see the average balance
in NOWs increasing faster than the average balance
in personal transactions accounts.

Turnover rate. A standard measure of the extent
to which an account is being used for transactions
purposes is the turnover rate of the account. The
turnover rate is the dollar value of transactions made
using an account in a year divided by the average
dollar balance in the account. It can be thought of
as the number of times per year an individual dollar
flows through the account-the higher the turnover
rate, the greater the transactions usage of the account.
For this reason, this measure is an important determinant of the monetary aggregate in which an asset
is included.
Table III compares the turnover rates of several
types of accounts. All the figures are actual values

As shown in Chart 1, the average balance in personal demand deposits has been fairly stable except
for the sharp drop following the authorization of
regular NOWs nationwide in 1981, as persons with
large checking account balances switched to regular
NOW accounts. Similarly, the average balance in
regular NOWs, although more variable, has been
stable except for the sharp fall after the introduction
of Super NOWs, as regular NOW holders with large
balances switched to Super NOWs. Finally, the
average balance in Super NOWs has been declining

Transactions Activity

12

The sharp fall in the average Super NOW balance
after the first quarter of 1984 seems to indicate that some
consumers initially consolidated savings and transactions
balances in Super NOWs but then transferred their
savings balances into MMDAs or MMFs. This could
have been caused by unusually high introductory rates on
Super NOWs.

Table III

TURNOVER RATES
Number of times
p e r y e a r1

Type of Account

Consumer demand deposits

18-30
15.8

NOW accounts
Money market deposit accounts

3.5

Money market mutual funds

2 . 62

Regular savings accounts

5.1

11

Since the growth in aggregate transactions and savings
balances reflects the behavior of both consumers and
businesses, it is possible that the slower growth of aggregate transactions balances is solely a result of the behavior of businesses. In this case, the aggregate data
would mask the fact that transactions and savings balances of consumers grow at roughly the same rate. If
this were true, however, the use of NOWs as savings
accounts by consumers would not alter the long-run relationship between the growth rates of M1 and GNP.

8

1
Annual averages of monthly data for 1984, except the turnover
rate of consumer demand deposits, which is an indirect estimate
for the early 1970s from Pugash (1974).

Excluding institutions-only funds;
Report of Holliston, MA 01746.

2

from

Donoghue’s

Money

Fund

Federal Reserve statistical release G.6, February 12, 1985,
Source:
except as noted.

ECONOMIC REVIEW, MAY/JUNE 1985

for 1984 except the turnover rate of consumer demand deposits, which is an indirect estimate for the
early 1970s from Pugash (1974).13 (Data are not
available to directly calculate the turnover rate of
consumer DDs). The turnover rate of NOWs
(including Super NOWs) is several times greater
than the turnover rate of savings-type accounts,
which implies that NOWs are used much more intensively for transactions purposes. The turnover rate
of NOWs is somewhat lower, however, than the estimated turnover rate of consumer demand deposits.
The lower turnover rate of NOWs than DDs is
not necessarily inconsistent with the view that NOWs
are primarily being used for transactions purposes
since an individual with a given amount of transactions would have a higher average balance if he.
were using a NOW than if he were using a regular
checking account. The two reasons for this were
discussed above. First, he would have to have a
higher average balance in order to satisfy the minimum balance requirement. Second, he would not
manage his transactions balances as closely in a
NOW as in a DD because the NOW pays explicit
interest. Since the turnover rate is calculated as the
dollar value of transactions divided by average account size, the individual’s higher average balance
in a N O W than in a DD (for the same a m o u n t o f
transactions) would cause his turnover rate to b e
lower. Therefore, the aggregate turnover rate of
NOWs would be somewhat lower than the aggregate
turnover rate of DDs even if NOWs were being used
solely for transactions purposes.
It is difficult to estimate just how much the higher

20 percent higher average balances is 15.8 X 1.2 or
19.0, which is within the estimated range of the turnover rate of consumer demand deposits. All in all,
the evidence on turnover rates is consistent with the
view that NOWs are being used primarily for transactions purposes.

minimum balance and lower opportunity cost of
NOWs would raise an individual’s average balance

Table IV

relative to what it would be in a DD for the same
volume of transactions. However, to illustrate the

Number of withdrawals and deposits. Two other
simple but direct measures of transactions activity
are the average number of withdrawals from and
deposits to an account. Although these measures
cannot be used to determine if consumers have consolidated their savings and checking balances in
NOWs, they can be used to determine if any NOW
accounts are being used purely for savings purposes.
Estimates of the average number of debits and deposits per account per month for personal checking
accounts (DDs), NOWs, and regular savings accounts are shown in Table IV. These estimates indiholding transactions balances was 2.35 percentage points
for Super NOWs, 4.51 percentage points for regular
NOWs, and 9.76 percentage points for regular checking
accounts, using the three-month Treasury bill as the
alternative asset. Based on an interest elasticity of the
demand for transactions balances of 10 percent, a representative individual would have held a 5.4 percent higher
balance in a regular NOW and a 7.6 percent higher balance in a Super NOW than he would have held in a
regular checking account. Combined! these two effects
would have caused a 17.0 percent higher balance in a
regular NOW and a 31.8 percent higher balance in a
Super NOW than in a demand deposit for the same
volume of transactions. Based on the relative amounts of
regular NOW deposits and Super NOW deposits, total
NOW deposits would have been 21.6 percent higher due
to these factors.

AVERAGE NUMBER OF DEBITS AND DEPOSITS
PER ACCOUNT PER MONTH
DEBITS

potential magnitude of these effects, assume that an

individual’s average balance is 20 percent higher in a
N O W .1 4 T h e turnover rate of NOWs adjusted for
13

It is difficult to assess whether the turnover rate of
consumer DDs has risen or fallen since this estimate was
made. Technological innovations have reduced the cost
of managing transactions balances. A t t h e s a m e t i m e ,
however, real wages have risen, thereby increasing the
opportunity cost of time spent managing balances. These
two factors would have opposite effects on the turnover
rate.
14

This 20 percent figure is plausible based on 1984 average data.
Average balances in regular NOWs and
Super NOWs were $4,826 and $12,844,. respectively.
Assuming that the minimum balance requirements were
$500 and $2,500, respectively, then minimum balances
alone caused a 11.6 percent higher average regular NOW
account balance and a 24.2 percent higher Super NOW
balance. Additionally, the marginal opportunity cost of

Personal DD

NOW

Regular
Savings

1981

16.68

14.63

0.36

1982

14.96

16.93

0.37

1983

17.19

15.66

0.42

DEPOSITS
Personal DD

NOW

Regular
Savings

1981

2.99

2.99

0.44

1982

2.62

3.46

0.47

1983

3.03

3.25

0.53

Source:

Functional

FEDERAL RESERVE BANK OF RICHMOND

Cost

Analysis,

1981-1983.

9

cate that the average NOW is just as active as the
average regular checking account and much more
active than the average regular savings account.1 5
They imply that there is not a significant number of
NOWs being used purely as savings accounts.
Seasonal Behavior
It is evident from a comparison of the behavior of
balances in regular checking accounts and passbook
savings accounts that transaction balances and savings balances have very different seasonal patterns.
This difference is illustrated in Chart 2, which shows
the 1984 seasonal factors for demand deposits and
passbook savings deposits.16,17 Demand deposits
increase strongly at times of seasonal transactions
needs such as the April tax date and the period before
Christmas. Savings deposits have a much weaker
seasonal movement, although they tend to decline at
the end of the year as people take funds out of their
accounts to finance Christmas expenditures.
15

The similarity in activity in NOWs and DDs is supported by the fact that banks’ noninterest expense per
account is nearly as high for NOWs as it is for personal
DDs. See Taylor (1984).
16

Despite financial innovation, the seasonal factors for
passbook savings deposits have been stable for the last
ten years.
17

The Board of Governors has not constructed explicit
seasonal factors for NOWs due to the data problems
mentioned below.

Chart 2

SEASONAL FACTORS
FOR DEMAND DEPOSITS AND
REGULAR SAVINGS DEPOSITS

If NOW deposits are transactions balances, then
their seasonal behavior should be similar to the seasonal behavior of consumer demand deposits. Unfortunately, there are two data problems in making this
comparison. First, there is a shortage of data with
which to evaluate the seasonal behavior of NOW
balances. The seasonal pattern of NOW deposits
was distorted by the introduction of nationwide
NOWs in January 1981 and by the introduction of
Super NOWs in January 1983. In addition, from
1981 through 1983 very strong growth in NOW deposits obscured the seasonal pattern. In 1984, however, there were no regulatory changes affecting the
intra-yearly pattern of NOW deposit growth and the
annual growth rate was only 11 percent. Hence, in
1984 it is possible to get a pretty good reading of the
seasonal behavior of NOW deposits.
The second data problem is that while NOWs are
held solely by consumers, demand deposits are held
both by businesses and consumers. Unfortunately,
there are no aggregate data available on consumer
demand deposits to compare to the NOW deposit
data. Until the end of 1978, however, the Federal
Reserve collected monthly data on gross demand
deposits at weekly reporting banks by type of holder.
We know from these data that the seasonal patterns
of consumer and business transactions deposits are
somewhat different.18 In particular, the buildup in
demand deposits in the months before Christmas is
greater for businesses than for consumers.
With these two problems in mind, Chart 3 compares the unadjusted monthly growth rates of NOW
deposits in 1984 to the unadjusted growth rates of
demand deposits and passbook savings deposits. The
strong similarity between the seasonal patterns of

demand deposits and NOW deposits-and the dissimilarity between the seasonal patterns of N O W
deposits and savings deposits-provides additional
evidence that NOW deposits are very heavily composed of transactions balances.

Feb.

Apr.

June

Aug.

Oct.

Dec.

One noticeable difference in the seasonal behavior
of NOW deposits and demand deposits occurs around
Christmas, when the seasonal movement in N O W
balances is smaller than that of DDs. This difference
might be taken as evidence that NOW deposits have a
significant savings component. It might, however,
simply reflect the different mix of deposit holders for
regular checking accounts versus NOWs. As noted
above, the buildup in demand deposits in the months
prior to Christmas-and the subsequent decline in the
months after Christmas-is considerably greater for

1984
Source: Federal Reserve Statistical Release, H.6.

10

18

See

Summers

ECONOMIC REVIEW, MAY/JUNE 1985

(1979).

with DDs had one of these accounts. In the aggregate, more NOW holders might use these alternative
accounts for seasonal transactions needs, rather than
building up balances in their NOWs.1 9 Second,
NOWs have higher minimum balances than consumer demand deposits. Both of these factors could
lessen the percentage movement in NOW deposits
around Christmas relative to that of consumer demand deposits.2 0

Chart 3

UNADJUSTED GROWTH RATES
OF NOWs, DEMAND DEPOSITS,
AND SAVINGS ACCOUNTS
Percent

SUMMARY

Feb.

Apr.

June

Aug.

Oct.

Dec.

1984
Note: Annualized growth rates of not seasonally adjusted
data less the trend rates of growth from December 1983 to
December 1984. NOW data includes demand deposits at
thrifts (see Appendix).
Source: Federal Reserve Statistical Release, H.6

business demand deposits than for consumer demand
deposits. Therefore, since only consumers hold
NOWs, one would expect a smaller amplitude in the
seasonal movement around Christmas of NOW deposits relative to that of total demand deposits.
Finally, even if NOWs are purely transactions accounts, there are reasons why the seasonal behavior
of NOW deposits might be somewhat different from
that of consumer demand deposits. First, NOW
account holders are more likely to have money market
deposit accounts (MMDAs) or money market fund
accounts (MMFs), which have some transactions
capabilities. In the 1983 Survey of Consumer Finances, discussed earlier, 24.6 percent of the respondents in the survey who had NOWs also had an
MMDA or a MMF while only 14.5 percent of those

This paper has examined two types of direct evidence on how consumers are using NOW accounts.
First, the survey responses by consumers on their
holdings of different types of accounts indicate that
few households have consolidated savings and transactions accounts into NOWs. Second, the characteristics of NOWs resemble those of regular checking
accounts and are very different from those of savings
accounts. Overall, the evidence provides little indication that a significant number of consumers are
using NOWs as savings accounts.
In conclusion, it appears that the introduction of
NOWs has not, in practice, weakened the distinction
between transactions accounts and savings accounts.
Consequently, there is little reason to believe that the
introduction of interest-bearing checking accounts has
significantly altered the long-run relationship between
M1 and economic activity or that M1 has deteriorated
as a measure of the public’s transactions balances.
19

A further implication of the relatively greater share of
NOW account holders with MMDAs or MMFs is that
total NOW balances might be more sensitive to changes
in market rates than total DD balances.
20

These two points raise the question of why the seasonal movement of NOW deposits around the April tax
date in 1984 was roughly equal to that of demand deposits. This probably reflects the different mix of deposit
holders for NOWs versus regular demand deposits. The
seasonal movement in transactions deposits in April is
largely due to the payment of nonwithheld federal in-ome
taxes by individuals. Since only consumers hold NOWs,
whereas both consumers and businesses hold regular
demand deposits, one would expect greater strength in
NOWs versus demand deposits in April compared to
other times during the year when there is a seasonal
demand for transactions deposits.

FEDERAL RESERVE BANK OF RICHMOND

11

APPENDIX

DATA SOURCES
Functional Cost Analysis The Functional Cost
Analysis (FCA) program is an annual survey by
the Federal Reserve of banks’ expenses and revenues.
The analysis is primarily intended as a management
tool for banks since it allocates costs and income
among the various functions of the bank and allows
the bank to compare its data to other banks. Participation is voluntary but substantial-608 banks in
1983. Average figures for all banks for each year
are made available to the public by the Fed.
The FCA data reported in the text (average size
of personal DDs; activity in NOWs, personal DDs,
and regular savings accounts) are averages for all
accounts of a given type in the survey. The personal
checking account category included both NOWs and
DDs, but figures for personal DDs could be isolated
based on the method in Taylor (1984). The key
assumption of this method is that the NOWs in the
personal checking category have the same characteristics as the NOWs in the NOW category.
Survey of Consumer Finances, 1983 Between
February and July 1983, 3,824 randomly selected
families were interviewed by the Survey Research
Center of the University of Michigan. The survey,
sponsored by the Board of Governors of the Federal
Reserve System and six other agencies, collected
data on families’ balance sheets. Information on the
use of NOWs is found in the section of the survey
on holdings of various types of assets. Detailed
information on the survey construction and on obtaining copies of the results is found in Appendix A
of “Survey of Consumer Finances, 1983” (1984).
NOW and Super NOW Account Size F o u r
times a year the Federal Reserve System surveys a
random sample of commercial banks stratified by size
on the number and dollar volume of NOWs, Super
NOWs, and MMDAs. The sample for NOW ac-

12

counts excludes banks in New England, while the
sample for Super NOWs and MMDAs is nationwide. The sample figures are used to calculate estimates of the number and dollar volume of these
accounts at all commercial banks. The average account size is calculated as the aggregate dollar volume
divided by the number of accounts. The estimates
are available upon request from the Banking Section
of the Board of Governors. In order to abstract from
the effects of inflation, the estimates of average account size in the text are divided by the personal
consumption expenditures deflator after first rebasing
it so that 1984 equals 100.
Rates Paid As of the last Wednesday of each
month, the Federal Reserve collects data from a
nationwide random sample of about 550 banks stratified by size on rates paid on Super NOWs and
MMDAs. Based on this survey, the Board estimates
average rates on these accounts at all commercial
banks. These estimates are published as part of a
special supplementary table to the H.6 statistical
release.
Turnover Rates The turnover rate is the ratio of
debits to deposits for an account type. The data are
reported in the Federal Reserve Bulletin and in the
Board’s G.6 release and, for MMFs, in Donoghue’s
Money Fund Report of Holliston, MA 01746.
Aggregate Deposits in NOWs As part of the H.6
statistical release, the Federal Reserve reports total
Other Checkable Deposits (OCDs), which consists
of deposits in NOWs (including Super NOWs) and
demand deposits at thrifts. Since demand deposits
at thrifts are less than 5 percent of total OCDs, the
aggregate data for OCDs were used in the text as
estimates of aggregate deposits in NOWs in order to
examine the seasonal behavior of NOW deposits.

ECONOMIC REVIEW, MAY/JUNE 1985

References
Corcoran, Patrick J., and Judy Wachtenheim. “NOWs,
Super NOWs, and the Interest-Elasticity of Household M1 Demand.” Federal Reserve Bank of New
York (June 1984), mimeographed.
Hafer, R. W.
“The Money-GNP Link : Assessing
Alternative Transaction Measures.” R e v i e w , F e d eral Reserve Bank of St. Louis (March 1984), pp.

Hetzel, Robert L. “The Behavior of the M1 Demand
Function in the Early 1980s.” E c o n o m i c R e v i e w ,
Federal Reserve Bank of Richmond (November/
December 1984), pp. 20-29.
J u d d , J o h n P . “The Recent Decline in Velocity: Instability in Money Demand or Inflation?” Economic
Review, Federal Reserve Bank of San Francisco
(Spring 1983), pp. 12-19.

Pugash, James Z. “The Demand for Money in Three
Sectors.” Board of Governors of the Federal Reserve System (January 1974), mimeographed.
Radecki, Lawrence J., and John Wenninger. “Shifts in
Money Demand : Consumers versus Business.”
Quarterly Review, Federal Reserve Bank of New
York (Summer 1983), pp. 1-11.
Summers, Bruce J. “Demand Deposits: A Comparison
of the Behavior of Household and Business Balances.” E c o n o m i c R e v i e w , F e d e r a l R e s e r v e B a n k
of Richmond (July/August 1979), pp. 2-14.
“Survey of Consumer Finances, 1983.” Federal R e s e r v e
Bulletin (September 1984), pp. 679-692.
Taylor, Herb. “ T h e R e t u r n B a n k s H a v e P a i d o n N O W
Accounts.”
Business Review, Federal Reserve
Bank of Philadelphia (July/August 1984), pp. 1323.

FEDERAL RESERVE BANK OF RICHMOND

13

DAYLIGHT OVERDRAFTS AND
PAYMENTS SYSTEM RISKS
David L. Mengle*

The last fifteen years have seen an almost incredible increase in the speed at which funds move through
the economy. Central to this change has been the
continual development of electronic funds transfer
(EFT) networks. Although few would deny the
benefits of the increasing facility with which transactions can take place, this progress has not come
without cost. Specifically, there has been increasing
concern with the risk exposure faced by participants
in the payments system, particularly with regard to
large dollar transfers that incur overdrafts.1 Along
with this concern have come proposals to deal with
this exposure.
On an average day in 1984, over $640 billion was
transferred by way of EFT networks. Given the
interdependence that exists between participants in
these networks, it is possible that losses in the event
of the sudden failure of an institution could be huge.
At the same time, however, such a failure is highly
unlikely, and has not in fact occurred. Thus, it is
possible for reasonable men to disagree on both the
magnitude of the problem created by overdrafts and
what to do about them.
The object of this article is to consider, in economic
terms, the nature of risks on EFT networks and the
desirability of specific measures proposed to deal
with these risks. With regard to the former, a simple
economic framework will be developed in Section II
in order to analyze risks by EFT participants. With
regard to the latter, Section III will describe and
evaluate the various policy alternatives that have been
advanced. The following questions will be implicit
in the discussion of risk reduction measures:
l

l

How will a risk reduction policy affect risk
assignment?
What incentives will the policy create among
participants?

l

Are the incentives created by the risk assignments likely to accomplish the policy’s objective of reducing excessive risk taking ?

Particular attention will be given to pricing as a risk
reduction policy and to the importance of the lender
of last resort to the problem of risk on private networks. This discussion will be followed by a description of policies adopted by the Board of Governors in
May 1985.
Since EFT systems are complex entities that are
little known outside the banking industry, Section I
will describe the major wholesale EFT networks and
explain the nature of the overdraft problem. Readers
who are familiar with EFT systems and daylight
overdrafts may wish to skip the first section and go
directly to the analytical material beginning with
Section II.
I.
A PRIMER ON EFT NETWORKS AND
DAYLIGHT OVERDRAFTS

Types of Networks
Generally, wholesale EFT systems are designed to
transfer funds between banks2 in order to permit a
customer (the “sender”) of the sending bank to
make a payment to a customer (the “receiver”) of
the receiving bank, or else to be used for payments
between banks. In contrast, retail systems, such as
automated clearinghouses or automated teller machine networks, are primarily, but not exclusively,
concerned with consumer payments. Wholesale wire
transfer systems may simply take the form of communications networks that convey instructions to the
receiving bank to debit the sending bank’s correspondent account and to credit the receiver’s account.
If the two banks do not have a correspondent relationship, the sending bank may instruct a third bank

* Economist, Federal Reserve Bank of Richmond.
1

General overviews can be found in David B. Humphrey
(1984), Richard L. Smoot (1985), and E. J. Stevens
(1984).

14

2

Throughout this article, “bank” will be used to refer to
any depository institution authorized to use EFT
services.

ECONOMIC REVIEW, MAY/JUNE 1985

having a relationship with both the sending and
receiving banks to debit the sending bank and credit
the receiving bank for the amount of the transfer.
These networks, of which BankWire and Society for
Worldwide Interbank Financial Telecommunications
(SWIFT) are examples, have in common the characteristic that they do not themselves provide settlement services, that is, they do not include any mechanisms for consolidating or centralizing transactions
between participating banks in order to determine
the banks’ financial positions in relation to each other.
As a result, wire transfers on nonsettling networks
are essentially bilateral exchanges.
Adding settlement services to a network characterized by bilateral transfers yields a gross settlement mechanism, in which each transaction is settled
when made. Because such a network uses a common
set of accounts, such as reserve accounts, to determine
relative financial positions, the receiving bank saves
the costs associated with maintaining and posting
entries to correspondent balances. In addition, since
settlement occurs immediately, the receiver could
have immediate access to “good” funds.
If all transactions between banks were conducted
on a bilateral basis, however, increasing numbers of
banks and customers in an economy would mean
increasing costs due to the sheer increase in volume
of transactions that must be handled separately. In
order to reduce such costs, participants in payments
systems have incentives to reduce costs by consolidating transactions into net debit and credit positions,
thereby reducing the number of actual interbank
transfers taking place. Specifically, net settlement,
in which transactions taking place within a specified
period are consolidated into net debits or credits for
each settling institution, allows each participant to
greatly reduce the number of payments and receipts
it must make per period. This can in turn take two
shapes : in its simplest form, net settlement means
that each pair of institutions comprising a payments
network would net final bilateral positions at the end
of, say, a day, and then arrange payment. Although
this would involve a payment or receipt each day
between a bank and all others with which it deals, it
does create economies by eliminating the necessity
for funds to change hands with each separate customer transfer. Further economies may be gained by
instituting “net-net” settlement, in which the settling
organization maintains a central network account,
and collects or distributes each institution’s position
vis-à-vis the settling organization at the end of each
period. As a result, each participant makes only one
payment or receipt per period. Because all existing

net settlement EFT networks use the latter form, net
settlement will be understood to mean net-net settlement in the remainder of this article.
Both gross and net settlement have their own
unique advantages. On one hand, gross settlement
provides immediate transfer and availability of funds.
On the other hand, net settlement has the greater
potential for cost reductions due to the consolidation
of payments and receipts. It is possible, therefore,
for both types of systems to exist side by side. When
choosing between systems for a particular transfer, a
bank would consider the importance of immediate
availability relative to other cost savings.
Large Dollar Wire Transfer Networks
Fedwire. The primary wholesale EFT network in
the United States today is Fedwire, operated by the
Federal Reserve System. This system, the first national settling network, has existed in various forms
since 1918.3 Settlement originally took place between
the twelve regional banks by means of the Gold
Settlement Account maintained at the Treasury, and
today is accomplished through the Fed’s Interdistrict
Settlement Fund. Fedwire uses a gross settlement
mechanism since, as will be shown, transfers between
participants are essentially bilateral, making use of
transfers between reserve accounts maintained at the
regional Federal Reserve Banks by the participants.
A transaction on Fedwire may take place as follows. A sender, who may be an individual or a
private or governmental organization, requests a bank
to transfer funds to a receiver. The sending bank
debits the sender’s account, and requests its regional
Federal Reserve Bank to send a transfer message to
the Reserve Bank serving the receiver’s bank. Banks
may be directly connected “on line” by computer
with the Fedwire system, or may use telephone or an
ad hoc computer link to make contact with their
Reserve Bank. The sending bank’s Reserve Bank
debits the sending bank’s reserve account for the
amount of the transfer, and credits the receiving
bank’s Reserve Bank for that same amount. The
receiving Reserve Bank debits the sending Reserve
Bank’s account, and credits the receiving bank’s
reserve account for the transfer. Finally, the receiving bank notifies the recipient of the transfer,
and the receiver is immediately free to use the funds.
This last point deserves emphasis. Fedwire transfers are final in that “irrevocable credit will promptly
3

For a description of payments mechanisms that evolved
before the Fed, see John R. Walter (1984).

FEDERAL RESERVE BANK OF RICHMOND

15

be given [by receiving banks] to the accounts of
customers receiving payments.“4 In other words, the
receiver may assume he has “good” funds as soon as
he is advised that a transfer has been made. Once
made, a transaction cannot be revoked. Further, if
the transfer takes place without sufficient funds being
provided by the sender, the Fed assumes the risk.
Thus, an attractive feature of Fedwire is the certainty
it provides to receivers.5
Although Fedwire services were originally offered
only to Federal Reserve member banks and were not
charged for, the Depository Institutions Deregulation
and Monetary Control Act of 1980 (DIDMCA)
instituted pricing of transfers and gave nonmember
banks direct access to Fedwire. The Fedwire system
has the highest number of transactions of any wire
transfer network. Because the Federal funds market
works through Fedwire, and immediately good funds
are desired in the money market, there is a continuing high demand for such a gross settlement
network. In 1984, approximately 166,410 transfers
took place on Fedwire on an average day, with an
average transfer size of $2.2 million each. Average
daily dollar volume was approximately $366 billion.
CHIPS. The second largest wholesale EFT system is the Clearing House Interbank Payments System (CHIPS), set up in 1970 by the New York
Clearing House Association. The original purpose
of the network was to clear international transactions
of member banks, but it now accommodates domestic
transactions as well. Settlement through CHIPS
takes place on a net-net basis, and is conducted by a
subset of participants known as settling banks. Approximately 138 banks participate in CHIPS, of
which 21 are settling banks. Of the remainder, about
two-thirds are branches of foreign banks.
A typical transaction may take place as follows. A
transfer may be initiated between 7 a.m. and 4:30
p.m. (New York time), during which period payment
information messages are sent from the sending bank,
through the CHIPS computer, to the receiving bank.
The computer in turn notifies the settling banks who
have agreed to settle for the sending and receiving
banks. This information is posted to the accounts
of the participating banks, but no settlement takes
4

49 Fed. Reg. 13190.

5

Transfers are certain due to a combination for finality
(Regulation J, Subpart B, 12 C.F.R. $210.36) and the
provision that the Fed will have a security interest in the
sending bank, not in the receiving bank or receiver, if an
overdraft is not covered by the end of the day (12 C.F.R.
§ 210.31).

16

place until after close of business. In other words,
funds transfers are provisional until settlement at the
end of the day.
After 4:30, the computer provides a list of net
credit and debit positions of the settling banks vis-àvis the CHIPS system and the position of the nonsettling banks against their settling banks. By 4:45,
this information should be provided to participants.
Once this becomes available, the settling banks may
notify nonsettling banks of their net positions. As
soon as all settling banks agree to settle, those in a
net debit position with CHIPS send payment,
through Fedwire, to the CHIPS account at the New
York Fed. Then, assuming all net debtors have
settled, payment is made to settling banks in a net
credit position with CHIPS by 6:00. At this time,
the CHIPS account should be back to a zero balance,
and all that remains are payments between settling
and nonsettling banks. Because Fedwire is used for
settlement within the CHIPS system, all net transfers
are final and there is certainty of the validity of the
transfers on the part of participating banks.
Finality of payment does not exist on CHIPS in
the same sense as it does on Fedwire. CHIPS transfers are irrevocable on the part of the sending bank,
that is, once sent, a payment remains an obligation of
the sending bank and cannot be cancelled. Unlike
Fedwire, however, the receiver’s account need not b e
credited until final settlement at the end of the day,
although the receiving bank is allowed to give
immediate credit if it so desires. Thus, transfers are
provisional rather than final until settlement occurs.
CHIPS has a smaller number of transactions than
does Fedwire, but a higher average transaction size.
For example, in 1984, almost 23 million transactions
took place at an average rate of over 91,000 transfers
per day. Average daily dollar volume was approximately $276.5 billion, and mean transaction size was
over $3 million.
CashWire. The third settlement network used in
the United States is CashWire, which began full
operations on April 1, 1952. This was developed as a
settling network from BankWire which, as pointed
out above, is a nonsettling network used to exchange
payments information and to effect transfers of correspondent balances.
CashWire is a net-net settlement system, and a
typical transaction occurs as follows. After a sender
notifies his bank that he wishes to make a transfer,
the bank transmits the payments information to the
BankWire computer, and the information is passed
through immediately to the receiving bank. At 4:30

ECONOMIC REVIEW, MAY/JUNE 1985

p.m. no more transactions are accepted, and a summary of net debit and credit positions is provided to
the banks and to the New York Fed. Fedwire is
then used by net debtor banks to send funds owed to
the CashWire account at the New York Fed. After
debit payments are received, Fedwire is again used
to remit funds to banks in net credit positions, the
CashWire account balance returns to zero, and payment is final.
CashWire has been characterized as providing
finality of payment in that receiving banks, in the
event of settlement failure, cannot take back funds
that have been made available to receivers.6 Actually,
transfers are apparently provisional to receivers until
net settlement, so finality on CashWire does not
exist for receivers any more than it does on CHIPS.
The closest CashWire comes to finality is its “receiver guarantee,” according to which banks must
make good the amount by which a failed sending bank
is unable to meet its net debit position. Specifically,
each of the failed bank’s creditors is required to make
up the shortfall by an amount proportional to its
share of the sum of all net credits extended on CashWire to the failed sending bank. This shortfall could
be made up from creditor banks’ liquid assets, but
nothing in the CashWire rules prohibits financing by
revoking funds that had been provisionally granted
to receivers. Thus, although finality may exist for
receiving banks in that a settlement, once computed,
will not be cancelled or unwound, it does not exist
for receivers until final settlement.
Cash Wire volume is small compared to Fedwire
and CHIPS, and daily volume as of August 1984
was only $500 million, consisting of 1,100 transactions per day on average.7 Participation has also
fallen short of expectations.8 As of Summer 1984,
out of 170 BankWire participants, only 22 were
members of CashWire as well.
In addition to the
Other wholesale networks.
foregoing, the Chicago Clearing House Association
operates the Clearing House Electronic Settlement
System (CHESS), which is open to institutions in
the Seventh (Chicago) Federal Reserve District.
The most recent addition to the list of networks is
the California Bankers’ Clearing House Association’s
Twelfth District Electronic Settlement System
(TESS) network for California banks in the San
6

See, for example, 49 Fed. Reg. 13190 and Stevens
(1984), p. 6.

7
8

American Banker, August 28, 1984.
American Banker, August 9, 1984.

Francisco Federal Reserve District. In Canada, the
Canadian Payments Association operates a system
that uses net-net settlement through the Bank of
Canada, while in England the Clearing House Automated Payments System (CHAPS) acts as a settling
network for a set of United Kingdom clearing banks.
Daylight Overdrafts
The major controversy regarding EFT networks
concerns the risk exposure of the payments system
due to the existence of daylight overdrafts. These
occur when payment is made during the course of a
business day before the transaction is covered with
“good” funds. In other words, at least one of the
institutions involved extends free credit that will be
repaid before the end of the day.
Daylight overdrafts on gross settlement networks
are not identical to those on net settlement networks.
On net settlement systems, overdrafts are of an ambiguous nature. Since participants on such systems
do not settle until the end of the day, the only overdrafts (in the strict sense) that occur are, first, between a sender who has not yet provided funds to
cover a transfer and a sending bank that allows that
transfer to go ahead immediately ; and, second, between a receiver and a receiving bank that allows the

receiver to use funds before settlement occurs. Both
of these may be thought of as normal credit judgments which banks are called upon to make. One
possible definition of a daylight overdraft on a net
settlement network, then, is the extent to which receivers have been allowed to draw on provisional
transfers. Another possible definition is the amount
by which a bank’s net debit position across all networks exceeds its reserve account balance with its
Federal Reserve Bank.9 In actual policy discussions,
however, daylight overdrafts on the private networks

are assumed to occur whenever a bank is in a net
debit position, regardless of reserve account balance.
Thus, by this definition, daylight overdrafts are an

inescapable result of the nature of a net settlement
system, in which at least one participant must be a
net debtor.
On a gross settlement network such as Fedwire, a
daylight overdraft has a more straightforward definition. Specifically, it refers to a transfer that has
been made and becomes final even though the sending
bank’s reserve account did not contain sufficient
9

The Canadian Payments Association uses a definition
If an
similar to this on its net settlement network.
institution’s net debit exceeds its clearing balance, it
receives an advance from the Bank of Canada which
must be paid back with interest.

FEDERAL RESERVE BANK OF RICHMOND

17

funds for the transfer at the time it was made. Alternatively stated, the sending bank’s reserve account
balance has gone negative.
On Fedwire, daylight overdrafts occur due to the
way in which Regulation J, Subpart B is written.
One section states that payment is final once the
receiving Reserve Bank sends it to the receiving bank
or else notifies the receiver of the credit. In addition,
this section makes it clear that finality means that the
receiver has the right to immediate use of the funds.10
Another section requires a sending bank to have
sufficient funds in its reserve account at the end of
the day to cover net debits for that day. 11 Thus, by
granting immediate use of transferred funds while
giving until the end of the day to cover debit positions, the opportunity for, as well as legality of, daylight overdrafts arises. Interestingly, the same section that gives the sending bank until the end of the
day to provide cover also empowers a Reserve Bank
to “refuse to act on” a transfer that “it has reason
to believe” may incur an overdraft, so that overdrafts
are apparently both permitted and frowned upon.1 2
It is not entirely clear how the regulation came to
be written so as to permit daylight overdrafts. Such
overdrafts probably occurred under the pre-1971
Fedwire system, which relied on manual accounting
and teletype notification of transfers. Since Reserve
Banks normally could not provide immediate information as to the current intraday status of a bank’s
reserve account, it is probable that wire transfers
took place before it was known that a sending bank
had sufficient covering funds in its reserve account.
During this period, there were no systemwide regulations specifically covering wire transfers, and any
rules that existed were contained in individual Reserve Bank operating circulars. Thus, it may be
argued that Regulation J, when it was finally rewritten to include wire transfers, simply formalized
what had already been taking place. However, when
the Board of Governors first proposed that Regulation J be revised to cover wire transfers, the relevant
section permitted transfers subject to the restriction
that “each transferer shall maintain . . . a daily net
balance sufficient to cover the transfers of funds
debited to its account.“1 3 Given this language, it is
10
11

1.2 C.F.R. § 210.36.
12 C.F.R. § 210.31.

not clear whether daylight overdrafts were to be
permitted or not. This was changed to the present
policy in the 1976 proposal, and the reason given was
“to clarify the amount of the balance which a member
bank must hold with its Federal Reserve Bank.“1 4
Indeed, given the technology in place at that time,
daylight overdrafts most likely could not have been
effectively controlled without major costly changes to
the system. In addition, reserve balances at that
time were higher than they have been since the
Monetary Control Act of 1980 imposed universal, but
lower, reserve requirements.
Once daylight overdrafts are permitted on a gross
settlement network, the distinction between a gross
and net settlement system begins to break down.
Although Fedwire’s transfers are final when made
while those on the other systems are not, sending
banks incurring overdrafts on Fedwire are allowed
to settle on a basis similar to that found on the private
networks. The distinction would be even less significant if participants in the networks were to grant
immediate irrevocable credit to their receivers.
II.
RISK AND WHOLESALE PAYMENTS NETWORKS

Risk Concepts
In order to gain some insights into the economic
aspects of payments system risks, assume there exists
a simple settlement network for banks in an economy
which permits overdraft transfers subject to their
being repaid by the end of the day, but does not guarantee that settlement will take place and does not
provide for finality of payment. This network may
use either gross or net settlement via a central set of
accounts. Banks undertake transfers for the benefit
of third parties (senders and receivers), and are
compensated by fees net of operating costs.15 Finally,
all payments are risky, that is, it is uncertain whether
or not an overdraft transfer will be covered. This
uncertainty means that costs may be borne by participants in the payments system. Such costs, known as
expected costs, are determined by multiplying the
magnitude of loss and the probability of such a loss
occurring.
Credit risk arises from uncertainty that funds
credited will actually be received, and is faced as a

12

Banks are not permitted to run overnight overdrafts,
and violators are subject to a penalty rate. S e e R e g u lation D, 12 C.F.R. § 204.7 and various Reserve Bank
operating circulars.
13

38 Fed. Reg. 32954 (1973).

18

14

41 Fed. Reg. 3098 (1976).

15

For a model that emphasizes the jointness of both
costs and benefits on payments systems, see William F.
Baxter (1983).

ECONOMIC REVIEW, MAY/JUNE 1985

private expected cost by the sending and receiving
banks and the receiver. This may in turn be broken
down into sender risk and receiver risk. In terms of
the simple payments system described above, sender
risk refers to the fact that a sending bank faces an
expected loss whenever it extends overdraft credit
to a sender. In other words, it is the risk that the
sender will not provide covering funds and is by no
means unique to EFT systems. At the other end of
the transaction, the receiving bank faces receiver
risk, which arises due to uncertainty whether or not a
sending bank will settle.16 In addition, if the receiving bank were to allow the receiver to draw on
provisionally transferred funds, the receiver also faces
expected costs due to the possibility that his bank
may attempt to revoke his funds in the event of
settlement failure.
Systemic risk refers to the expectation that failure of one bank to settle will cause another bank or
banks to fail to settle as well. This would arise if,
within a particular bank, ability to settle debit positions depends on receipt of credits. Default of a
particularly large net credit (in relation to total liquid
assets) may keep a bank from meeting its obligations
to other banks against which it is in a net debit position. In turn, this failure could conceivably cause
settlement failures at other banks who depend on
credits from the receiving bank to meet their debit
positions, and so on. Thus, systemic risk refers to
expected costs that are not borne solely by the incurring bank but by other participants in a payments
network as well. In other words, a receiving bank
that accepts a transfer and allows a receiver to draw
funds before settlement incurs not only private costs
due to credit risk but also external costs that are
borne as receiver risk by other banks in the system.
These latter costs, known as externalities, may not be
borne exclusively by the receiving bank’s creditor
banks, but also by banks to whom these creditors are
in overdraft positions. Thus, in accepting a transfer,
a receiving bank can be expected to take into account
its private credit risks but not these additional social
costs. The end result in the simplified payments
system is an incentive for receiving banks to accept
riskier transfers than would be the case if these banks
were to bear all their costs privately.
The distinction between private and external costs
is important for the following reason. Since receiver
risk is borne privately by a receiving bank, this bank
may be expected to have incentives to keep such costs
16

Stevens (1984) combines sender and receiver risk into
settlement risk.

to a minimum. Specifically, a bank may wish to
control its exposure to sending banks it has reason
to believe may default on credit positions. However,
systemic risk is borne by other banks in the system,
so that this same bank may not have incentives to
limit risks to which it exposes other banks. In other
words, there is little reason to expect this bank to
place as much emphasis on controlling its net debit
position against the rest of the system as it would
place on controlling its exposure as a creditor to
other banks.
In this simple system, it should make little difference for risk purposes whether a network uses net or
gross settlement. Under the former, sender risk
would exist until an overdraft is covered, while receiver risk would exist until final settlement. Under
the latter, both sender and receiver risk would exist
until the overdraft is covered. Assuming all such
transactions under either system must be covered by
the end of the day, potential risk on net settlement
networks is the upper limit for that on gross settlement networks.
Risk Assignment Under Varying
Institutional Structures
Both CHIPS and CashWire have some similarities
to the hypothetical payments system described above.
There is no finality for receivers, nor is there explicit
guarantee of settlement to receiving banks. Thus, all
the risks found in the hypothetical model are also
found in the private networks. Sender risk exists for
those banks sending transfers for customers who have
not provided cover at the time of transfer. Receiver
risk exists until final net settlement occurs at the end
of the day. If the receiving bank allows a receiver to
draw funds before settlement, it is exposed to risk
even though the receiving bank may try to recover
the funds from the receiver. In other words, there
is no finality of payment to receivers until final net
settlement. Thus, receiver risk is borne by both
receiving bank and receiver. Less directly, the sender
may also be at risk because, if a receiving bank successfully recovers funds from a receiver after failure
of a sending bank, the receiver may have cause for
action to recover payment from the sender. In turn,
this sender may have already supplied funds to the
failed bank.
Systemic risk is present on this simplified version
of the private networks because a bank may depend
on a credit from a failing bank to pay other banks
against which it is in a net debit position. If a

FEDERAL RESERVE BANK OF RICHMOND

19

CHIPS member fails to settle, and no other bank
will settle for that member, all debit and credit messages for this sender may be cancelled and a new
settlement computed. Systemic risk arises here because of possible dependence of other banks on credits
from the failed bank. The cancellation of the messages does not relieve the failing bank of obligation
to ultimately settle because CHIPS transfers, once
sent, are irreversible. However, this does not affect
systemic risk since it is highly improbable that funds
could be recovered during whatever remains of the
day before settlement. On CashWire, as noted above,
losses are apportioned among creditors. Although
this does not by itself eliminate systemic risk, it does
appear to localize problems.1 7
Adding finality of payment, under which immediate and irrevocable credit is granted to receivers, to
the hypothetical system varies the risk assignment
somewhat. Sender risk does not change, but receiver
risk, which was previously shared by the receiver
and the receiving bank, is narrowed to the receiving
bank. One would expect this new risk assignment
to lead to incentives on the part of receiving banks to
monitor the soundness of sending banks from which
they receive transfers. At the same time, receivers
(and, as will be shown, senders), will have fewer
incentives to monitor sending banks.
Adding guarantee of settlement to finality adds an
additional party, the insurer, to the hypothetical payments system model, thereby approximating the risk
assignment on Fedwire. Finality removes risk from
the receiver, thereby confining receiver risk to the
receiving bank. However, the Fed, as insurer, guarantees that settlement will occur, that is, that the
overdraft will be covered. In the event of settlement
failure, the Fed has recourse to the failing bank,1 8
although the amount eventually recovered is uncertain. Thus, receiver risk is assumed by the Fed and
consists of expected losses net of amounts expected
to be recovered from the bank that failed to settle.
Because any loss would reduce Fed net revenues
available for transfer to the United States Treasury,
the risk is ultimately borne by the public. Sender
risk on Fedwire does exist, and may be borne by
the sending bank and the Fed. However, since
receiver risk is shifted from the receiving bank to the
public, failure to settle is localized and will not affect
17

Nothing in the CHIPS rules appears to preclude the
New York Clearing House Committee from dealing with
settlement failure in the same manner as that specified
for CashWire.
18

12 C.F.R. §210.21.

20

the position of any other bank. Still, it is important
to emphasize that no receiver risk has been eliminated; rather, it has simply been socialized.
The question of what happens to systemic risk is
more complex. If banks no longer face receiver risk,
they cannot face systemic risk so, at first blush, it
appears that systemic risk has been eliminated. Certainly the externality has not been placed on the
participating banks, since banks are not made to take
these costs into account when choosing whether to
accept a transfer. However, from the point of view
of the insurer, that is, the Fed, systemic risk is someone’s receiver risk. This is analogous to the fact that
externalities are the sum of private costs borne by
individuals other than those incurring them. For
example, when a factory causes pollution, it inflicts
costs on nearby landowners. The sum of these costs
is the externality incurred by the polluter. If the
factory owner is made to compensate the surrounding
landowners for the pollution he has caused, all the
pollution costs are turned into private costs to the
polluter. Similarly, the total risk assumed by the
insurer is the sum of receiver risks in the system,
and in assuming all receiver risks the Fed has thereby
assumed systemic risk as well.
Insurance of receiver risks in the Fedwire system
means that, other things equal, costs faced by banks
when exchanging payments messages are lower than
they would otherwise be without insurance. As a
result, supply of messages is increased. In other
words, since banks need no longer concern themselves
with receiver risk, they may tend to accept transfers
from sending banks who may have been turned down
if no insurance had been provided. Thus, the Fed
provides a valuable service to each bank by assuming
receiver risks. Since the Fed’s insurance is provided
at zero price, banks have little incentive to reduce
exposure to overdraft transfers.
It may also be argued that the Fed also provides
implicit insurance on the private networks through its
role as lender of last resort. According to this line
of reasoning, the Fed would never sit back and allow
systemic failure, and would surely step in by lending
to banks in net credit positions with a failed bank in
order to contain the effects of any settlement failure.
If this is indeed the case, then the final assignment
of costs depends on what happens to the receiving
banks affected by a sending bank’s failure. Here, it
is important to distinguish between illiquidity and insolvency.
If a receiving bank’s problem is insufficient liquid
assets to make up the sending bank’s shortfall, discount window lending to the receiving bank in order

ECONOMIC REVIEW, MAY/JUNE 1985

to allow settlement to proceed would have the immediate effect of eliminating systemic risk. However,
since the receiving bank would have to pay back the
loan plus interest to the Fed, that bank would end up
bearing the receiver risk net of any amounts eventually recovered from the failed sending bank. Further,
the Fed will be compensated for the credit it has
extended to prevent failure. Finally, the externality
cost will be placed on the borrowing bank, since by
averting failure due to illiquidity, the borrower is
prevented from passing costs on to its creditor banks.
Thus, if failure of one participant in a system causes
liquidity problems for its creditors, discount window
lending to the failed bank’s creditors will both prevent
systemic failure and assign costs to receiving banks
and receivers. This in turn would increase incentives
to monitor sending banks.
Insolvency presents a more complex set of circumstances. If a bank in a net debit position fails without
warning before net settlement occurs, it is possible
that one or more receiving banks may eventually
become insolvent as a result. In this case, the assignment of costs will depend on to whom the Fed lends.
If the Fed advances credit to the failing receiving
banks in order to allow settlement to proceed, then
the costs will ultimately be borne by the Federal
Deposit Insurance Corporation fund and the public.
In this case, the external costs are not placed on the
failed banks, although the discount rate represents a
price of the Fed’s assuming the risk of having to bear
these costs. If the Fed only advances credit to banks
experiencing liquidity problems as a result of some
receiving banks failing, then the assignment of costs
is identical to that described in the previous paragraph.
It is important to emphasize that the preceding
analysis depends on the assumption that the Fed
lends to receiving banks in net credit positions with
the failing bank. If, on the other hand, discount
window advances were made to the failing bank
simply to allow settlement to proceed, costs would,
as in the case of insolvency, be shifted to the FDIC
and public and not to the receiving banks. In this
case, as with Fedwire, receiving banks would have
little incentive to monitor sending banks. Thus,
banks’ incentives to control risk exposures may be
related to their perceptions of to whom the Fed is
likely to lend in the event of settlement failure.
To sum up, the crucial difference between risk
assignment on Fedwire and that on the private networks stems from the manner in which transfers are
guaranteed. On Fedwire, losses due to failure to
settle are borne by the public free of charge. There-

fore, since sending banks are not made to take account of external costs involved in making transfers,
there is an oversupply of transfers. On the private
networks, however, risks are placed on receiving
banks because banks will have to repay the lender of
last resort the amounts borrowed to cover a sending
bank’s failure to settle. Since interest is charged for
this lending, the externality should be passed back to
the banking system, and should be faced as a cost by
banks when accepting transfers from each other.

III.
POLICY RESPONSES

The complexity of payments systems institutions,
to say nothing of the overdraft problem itself, makes
it critical that any policies instituted to control risk
be selected thoughtfully and deliberately. Using the
framework developed above, this section will analyze
five policy alternatives. Two of these, pricing and an
intraday funds market, explicitly rely on the price
system to reduce risks. Two others, banning daylight
overdrafts and placing restrictions on overdrafts,
explicitly reject price incentives. The fifth, finality
of payments, creates an assignment of liabilities
among parties to a transfer in order to provide incentives to monitor risks. All five will be looked at in
terms of how they affect risk assignment, what incentives they will create, and whether they may be
reasonably expected to attain their stated objectives.
Following this discussion, recent policy initiatives
from the Board of Governors will be described.
Analysis of Policy Alternatives
Ban daylight overdrafts. As was pointed out in
the first section, daylight overdrafts on net settlement
systems actually refer to net debit positions. Since
if one or more parties on a net settlement network
are net creditors, then at least one other party must
be a net debtor. It follows that banning daylight
overdrafts on private networks would mean that a
bank could incur a debit only if it were receiving an
offsetting credit in return. Such a ban would, in all
likelihood, be so costly as to eliminate net settlement
systems entirely.
Daylight overdrafts could be banned on a gross
settlement network. If daylight overdrafts on Fedwire were totally forbidden, so that transfers could
only be made if banks had sufficient funds in their
reserve accounts to cover them, sender risk would
still exist to the extent that sending banks extend

FEDERAL RESERVE RANK OF RICHMOND

21

overdraft credit to their sending customers. However, receiver risk to the public would be eliminated.
Thus, from a risk reduction standpoint, banning overdrafts would be the most effective course of action,
at least as far as Fedwire is concerned. Unfortunately, such a policy would be fraught with difficulties.
Forbidding Fedwire overdrafts would be costly to
the banking system and its customers because transfers would have to be held until covering funds were
provided, thus depriving institutions of flexibility in
making transfers. The results would be the intraday
analogue of forbidding short-term credit by which
businesses bridge gaps between payments and receipts. Of course, this is not to argue that the current
level of overdrafts is somehow optimal, but rather
that some overdrafts may be justified on efficiency
grounds. 19 If daylight overdrafts are permitted, however, there is no reason why they should be given
away as free credit.
From an operational standpoint, banning daylight
overdrafts would be costly if one were to insist on
“real time” (second-to-second) monitoring of Fedwire transfers in order to stop overdrafts before they
occurred. A less costly approach would be to monitor
transfers ex post and then to levy heavy penalties in
order to deter would-be violators. Finally, a strict
ban on daylight overdrafts on Fedwire could easily
be evaded by shifting transfers to one of the private
networks.
Establishment of an intraday credit market. Daylight overdrafts could eventually be banned if an
intraday Fed funds credit market were to evolve.
There, credit would be available for periods of less
than 24 hours, possibly by lending for four-hour
increments. If such a market existed, a sending bank
could make the decision whether to borrow in order
to send immediately or else to wait until covering
funds were on hand. This would preserve the flexibility of the present system, but would shift risk to
intraday lenders rather than to the Fed. Further,
since risk would be borne by lenders only for a price,
the costs described above would be placed back on

19

For example, it has been argued that banks in unit
banking states must depend on daylight overdrafts more
heavily than do banks in other states. (Chicago Clearing
House (1984), Appendix A) In addition, many overdrafts are apparently the result of current practices of
banks buying and selling federal funds. To the extent
that these practices simply reflect institutional practices
that evolved as the result of permitting overdrafts, they
do not affect the arguments presented here.

22

those banks that incur them.20 Most importantly,
since intraday credit would no longer be unpriced,
borrowers would have incentives to economize on
risks they incur. Since effects of such a system on
resource allocation are similar to pricing, further
discussion will be suspended until the following
section.
Pricing. An alternative perspective on the overdraft problem can be obtained from the economics of
information and uncertainty. 21 Microeconomic theory
asserts that, given a choice between, say, receipt of
$200 for certain and a fifty percent chance of $1000,
a rational person may prefer the certain $200. In
other words, even if this person could expect to win
$500 in such a gamble on average, he may be willing
to give up some expected winnings in order to reduce
uncertainty. This behavior is referred to as risk
aversion, and one of its implications is that riskaverse individuals are willing to pay to have uncertainty reduced, that is, they are willing to buy
insurance at some price.
Assuming that the stockholders of banks participating in a payments network are risk-averse, it
follows that they would be willing to pay a positive
amount to have receiver risk reduced. In other
words, insurance of receiver risk is a valuable service
to banks. Thus, the Fed’s nonpriced guarantee of
Fedwire overdraft transfers is in effect a subsidy to
the stockholders and customers of participating
banks. The amount of the subsidy is not the actual
amount of overdrafts, but rather the premium that
bank owners would be willing to pay to have receiver
risk assumed by the insurer. If this insurance were
explicitly priced, not only would the subsidy be recovered but banks would have incentives to take
account of the risks they place on the payments
system. Thus, although credit risk would still be
assumed by the public, it would not be assumed for
free.
One advantage of pricing Fedwire overdrafts is
that it acknowledges that a certain amount of overdrafting may be optimal insofar as it helps maintain
an “efficient” payments system that avoids a gridlock
in which the whole system becomes jammed due to
delayed transfers. Overdrafts could be monitored
on an ex post basis and then charged for, say, at the
end of the month. Banks would, as a result, be made

20

Cf. Kenneth J. Arrow’s (1969) characterization of the
externality problem as “a special case of a more general
phenomenon, the failure of markets to exist.”
21

See, for example, John D. Hey (1979).

ECONOMIC REVIEW, MAY/JUNE 1985

aware that they impose costs on the system, and thus
would have incentives to delay at least some transfers
until covering funds become available. In addition,
charges made on the basis of percentage of overdrafts
would bear some relation to risk exposure, certainly a
closer relationship than that found in the current
practice of charging a flat fee for wire transfers
regardless of overdraft position.
Pricing of overdrafts on Fedwire may also be
desirable from the standpoint of public policy regarding competition between payments service providers.
Since Congress passed the Monetary Control Act of
1980, the Federal Reserve System has been placed in
the somewhat awkward position of both regulator and
competitor of private banks. Although the Act contains no specific mention of “fair” competition with
the private sector, the legislative history of the Act
shows that Congress clearly was concerned with this
subject. In fact, some have recommended that the
Federal Reserve Act be amended to make such competition an explicit objective of Fed policy. 22 Despite
the ambiguity of current law, however, the Board of
Governors has expressed its “fundamental commitment to competitive fairness” and stated as a matter
of policy that “Federal Reserve actions are . . . implemented in a manner that insures fairness to other
providers of payments services.“23 Thus, since guarantee of settlement in the form of free intraday credit
constitutes a competitive advantage of Fedwire over
the private networks, pricing may help to stimulate
competition in the provision of payment services.2 4
The major obstacle to implementing a pricing
scheme on Fedwire is that it is not obvious what the
appropriate price for either insurance or intraday
credit is. Risk premia most certainly exist in a world
of risk-averse individuals, but measuring such premia
would require data that are typically not available,
such as how much risk bank stockholders are willing
to tolerate and the probability of settlement failure
occurring. A market price for intraday credit does
exist in the form of the broker call rate charged brokers for day loans, which are made so that a broker
can certify a check used to pay for securities. This
22

See Raymond Natter, “Legislative Intent Regarding
Pricing of Services by the Federal Reserve Board” in
U. S. Congress, House of Representatives (1984), pp.
81-91; for the recommendation, see p. 8.

rate tends to float about 100 basis points above the
Fed funds rate, and the loans are repaid by the end
of the day.2 5 The main problem with this form of
pricing is that it charges a one-day rate whether the
actual loan lasts a few hours or the whole day. There
is no reason, however, why the rate could not be
computed for shorter periods. An alternative means
of pricing is to use the tax-adjusted difference between the rates on go-day bank certificates of deposit
and Treasury bills as an approximation of the price
of default risk.26 The problem with this approach is
that this difference fluctuates widely, and disentangling default risk from tax effects is likely to be a
formidable task.27 As another alternative, overdrafts could be charged the Fed funds rate extrapolated backwards to periods of less than 24 hours.
If such a rate is too high, then it is likely that an
intraday credit market would develop. If it is too
low, at least it will provide incentives in the desired
direction, that is, toward fewer overdrafts. Finally, a
price for overdrafts could be computed from the discount rate. Since this rate is typically lower than the
Fed funds rate, it is less likely to lead to an intraday
funds market.
While pricing daylight overdrafts may be desirable
on Fedwire, it does not necessarily follow that net
debits on the private networks should be explicitly
priced as well. As pointed out above, if the Fed
stands ready to lend through the discount window in
order to prevent systemic failure due to illiquidity of
receiving banks, the costs incurred by the failed bank
will most likely be borne by the banks in net credit
positions against the failed bank. Since these costs
represent expected costs to participants in net settlement networks, banks should take them into consideration when deciding on risk exposures. Thus,
although explicit pricing of risks does not exist on
the private networks, there is implicit pricing so long
as banks expect the Fed to lend to net creditor banks
experiencing liquidity problems due to the failure of a
net debtor to settle. In addition, although the public
may bear some losses if one or more of the receiving
banks fails and is unable to repay borrowed funds,
the Fed is compensated for taking this risk because
interest is charged on discount window credit. The
main problem for policymakers is to ensure that
sufficient collateral is on hand to facilitate discount

23

Board of Governors of the Federal Reserve System
(1984), pp. 710, 712.

24

The Department of Justice has also expressed its concern with such an advantage, and has suggested consideration of pricing. See U. S. Department of Justice
(1984), pp. 7, 34-5.

25

For a discussion of day loans, see Gardiner B. Van
Ness III (1975), pp. 143-52.

26
27

Humphrey (1984), pp. 100-l.
See Timothy Q. Cook and Thomas A. Lawler (1983).

FEDERAL RESERVE BANK OF RICHMOND

23

window lending should systemic failure be threatened.2 8
The existence of implicit pricing on the private
networks calls into question the assertion that certain
intervention by the lender of last resort creates a
“moral hazard,” that is to say, leads network participants to devote fewer resources to monitoring the
riskiness of sending banks than would otherwise have
been the case, thereby increasing the probability of
settlement failure.29 For moral hazard to exist, however, it would be necessary for banks to be able to
shift the costs of their failure to monitor to other
banks. At least in the case of temporary illiquidity
of receiving banks, it is difficult to see how such cost
shifting could occur, since borrowers must pay back,
with interest, funds advanced by the Fed. Thus, if
banks are not observed to engage in extensive risk
monitoring, it may be due not to moral hazard but
rather to their perception of a very low probability of
settlement failure.
Finality of Payment. As mentioned above, finality
exists on Fedwire but not on CHIPS or CashWire.
In essence, finality of payment establishes a strict
liability rule under which a receiving bank is made
liable for all payments it accepts. Regardless of
whether or not the receiving bank could have foreseen
the failure of a sending bank, the receiving bank
would have no recourse to the receiver. The rationale
for such a condition is that overall costs would be
minimized by focusing them on the party to the
transaction who can reduce risks at lowest cost.30 In
other words, the receiving bank is made to bear the
costs of a settlement failure because it is believed
that this party is in the best position to monitor and
avoid such costs. Although finality would not by
itself reduce risks, it has been justified as a means of
risk concentration that would in turn induce banks
to take risk reduction measures, thereby minimizing
costs to all parties.3 1
28

Institutions that do not have access to the discount
window could be required to collateralize all daylight
overdrafts, or else these institutions could be forbidden
overdrafts.
29

Stevens (1984), p. 11.

30

This corresponds to the “cheapest cost avoider” in
Guido Calabresi (1970), pp. 135-40.

It is possible, however, that the receiving bank
may not necessarily be in the best position to monitor
sending banks, since it would require each bank to
perform a credit evaluation of each participant in the
system. Although rating firms exist to evaluate
creditworthiness, it may be costly to obtain continually updated ratings. Thus, it may be preferable to
place some liability on senders and receivers as well.
There would be at least two advantages to such an
assignment. First, customers using large dollar
transfer networks to send funds are likely to possess
the sophistication to monitor the banks with whom
they deal. Second, they would, as customers, probably have fewer banks to monitor than would banks
participating in a funds transfer network.
If a sending bank fails before net settlement on a
network with no finality, a receiving bank that had
allowed a receiver to use funds before they were
finally received may now attempt to take back those
funds. If the attempt is successful, the receiver now
has reason to take action against the sender to recover a payment. This exposes the sender to liability.
This is especially severe since that sender may have
actually provided funds to the failed sending bank.
If he had, then he must both attempt to recover from
the failed bank and also will be subject to action by
the receiver demanding payment. Thus, the sender
will have incentives to monitor the riskiness of the
sending bank he selects. At the same time, due to the
uncertainty of recovery from the receiver by the
receiving bank, there still will be incentive for this
bank to monitor the riskiness of sending banks.
Finality of payment, then, is justified only if it can
be shown to be the cost-minimizing assignment of
liabilities. If monitoring of sending banks is costly,
then it may be preferable to spread liability in order
to give other parties incentives to monitor. The
benefits of finality become even more dubious when
one considers that lack of recourse to transferred
funds may increase the probability of settlement
failure for a receiving bank, and thereby increase
systemic risk as well. Although finality may “insulate
the nonbank sector from the effects of a settlement
failure,“ 3 2 it is not at all clear that it creates the
incentives that would minimize risks from a social
standpoint.

31

See Stevens (1984), pp. 10-11. Stevens also suggests
that a “hands-off” policy by the Fed toward settlement
failure would create incentives similar to those claimed
for finality.

24

32

49 Fed. Reg. 13190.

ECONOMIC REVIEW, MAY/JUNE 1985

Nonprice rationing of overdrafts. Approaches to
payments system risk that ration daylight overdrafts
seek to control either a bank’s exposure to risk from
sending banks or else the amount of risk one bank
creates for the rest of the payments system. Indeed,
such measures have been recommended by private
sector studies.3 3
Net bilateral credit limits are drawn up by a receiving bank and specify the maximum net transfer
the bank will receive from a particular sending
bank.34 By limiting the size of a net transfer, exposure of a receiving bank to a sending bank is kept
within bounds. Since failure of a sending bank would
inflict costs on a receiving bank, it is likely that, if
bilateral limits were an effective risk reduction measure, banks would institute them. Further, if finality
of payment were imposed on a system, receivers may
have greater incentives to establish bilateral limits.
Indeed, all private payments systems now have such
limits in place.3 5 However, such a measure is not
without problems.
The most serious problem with bilateral credit
limits is, as in the case of finality, the cost of making
judgments about individual banks in order to set
actual limits. Judgments would include determining
for which banks to set limits, gathering information,
analyzing information, and updating the limits as
conditions change. Since many banks would be faced
with large numbers of judgments to make, some sort
of categorization may be necessary in order to obviate a separate study for each sending bank. However,
this would involve a loss of detailed information,
which would in turn make the resulting bilateral
limits less useful.
A second problem with net credit limits is that
they, like most nonprice rationing schemes, tend to
be inflexible and therefore costly. Suppose a receiver
is expecting a payment over a private network, but
that the payment exceeds the receiving bank’s net
credit limit. If limits were rigid and could not be
easily modified, they would preclude the bank from
accepting the payment, even if there were no doubt as
to the sending bank’s solvency. 36 In actual practice,
33

See, for example, Association of Reserve City Bankers
(1983), pp. 23-25.

34

49 Fed. Reg. 13189.

however, limits can be lifted to cover such situations,
The result is that flexibility is preserved but the value
of the limits as a risk reduction tool may be called
into question.
The third problem with bilateral limits is intimately related to the second. If limits are set at relatively low levels, they may reduce risks but may also
send business over to Fedwire, where no such limits
exist. This problem would be even more acute if
finality were imposed on private networks. Thus,
the problem of competitive equity between public and
private networks again rears its head.
Net debit caps attempt to control the risk a bank
poses to the banking system by limiting the amount
by which a bank may be “in the red” on a network
or across networks.37 Such a measure should be
effective for reducing both receiver risk and systemic
risk. Unfortunately, such caps have the same disadvantages of inflexibility as do net credit limits, although they do not involve as costly a set of information requirements to implement. In addition, since
net debit caps would control costs external to individual banks, it is unlikely that these banks have
strong incentives to establish binding caps that are
likely to limit risks to acceptable levels, Thus, such
caps, if adopted, would probably have to be developed
by a collective effort of banks or, failing that, imposed
by regulatory fiat.
Net bilateral credit limits and net debit caps may
be characterized as measures which limit risks by
limiting the choices of banks without altering the
underlying incentive structure. In other words,
neither may be expected to affect risk assignments
among network participants, but rather to control
the amount of risk assumed. Because both may
prevent some transfers from taking place, they are
likely to reduce risks in the short run. Over the long
run, net bilateral credit limits are likely to prove
useful to banks as means of controlling exposure to
other banks in the system. However, it may be in
some institutions’ interests to find ways to circumvent such restrictions as net debit caps, thereby
short-circuiting risk control policies. For that reason,
it may be necessary in the future to institute additional policies that more directly affect the incentives
of payments network participants.

35

CHIPS was the last system to institute bilateral net
credit limits. American Banker, December 5, 1984.

36

A likely response is to divide the payment between two
networks.

37

49 Fed. Reg. 13188. CashWire has net debit caps of
50 percent of capital, while CHIPS is experimenting
with a more complex system of caps.

FEDERAL RESERVE BANK OF RICHMOND

25

Recent Policy Initiatives

IV.

As mentioned above, all private networks now
have bilateral net credit limits in place. In October
1984, the Association of Reserve City Bankers
(ARCB) Risk Control Task Force issued a report
outlining a procedure for setting up a voluntary
system of net debit caps.3 8
The caps would work as follows. Financial institutions would evaluate themselves in the areas of
creditworthiness, operational controls and procedures,
and credit controls. The ratings in these three areas
will then be combined in order to give an overall
rating-exceptional, satisfactory, or less than satisfactory. The rating would determine whether a bank
would be permitted to overdraft across systems up to
2.5, 2, or 1.5 times capital, respectively. However,
the limit would not apply to individual overdrafts but
rather to the average maximum overdraft over a twoweek reserve period. Finally, the caps would be
applied across: all wholesale EFT networks, including
Fedwire.
In May 1985, the Board of Governors met to discuss risk reduction measures. Generally, they agreed
with the ARCB recommendations, but instituted two
major changes to the net debit cap system.39 First,
in addition to a cap on average maximum overdrafts,
banks will be asked to set higher caps on maximum
daily overdrafts. Controls will be applied by banks
to both average overdraft size and how much these
overdrafts will be allowed to vary from the average.
Second, average caps were reduced by fifty percent
of capital for each rating category, so that a bank
rating itself in the highest range could not overdraft
more than twice its capital. The Board also added a
fourth, lowest category, the members of which could
not overdraft at all.
The regulators’ role in this self-regulatory mechanism would be for examiners to review the selfevaluations and to point out areas of disagreement.
Such a role is actually compatible with a resultsoriented approach to regulation, in which specific objectives are established by regulators but implementation is left to the regulated industry. The older,
alternative approach would be for the regulator to
require specific actions to accomplish the objectives.
This latter approach would only be brought into play
if the former approach fails.
38

Association of Reserve City Bankers (1984).

CONCLUDING COMMENTS
On its surface, the risk reduction problem has all
the earmarks of the economic problem of collective
action, that is, actions that are in the interest of
depository institutions as a whole are not necessarily
in their interests as individuals. If this accurately
describes the situation with regard to systemic risk,
then the current “voluntary action in lieu of regulation” solution is warranted. Because the alternative
to voluntary action is more stringent imposed regulatory solution, the banking industry may well have
incentives to attempt to reduce overdrafts by means
of a self-regulatory mechanism.
However, this article has attempted to demonstrate
that while the problem is complex, the solution need
not be. On Fedwire, risk is being assumed by the
public because intraday credit is granted free of
charge to participating depository institutions. Thus,
some sort of pricing of overdrafts or an intraday
funds market may be called for. Even with pricing,
Fedwire would continue to be in demand because it,
unlike the private networks, provides immediate
transfer of funds, On the private networks, however,
pricing is not called for because risks are borne by
participating banks. If the Fed lends to receiving
banks through the discount window in order to prevent a systemic failure, the borrowing institutions
will bear the cost, and may be expected to take this
cost into account in their credit decisions. Further,
even without finality of payment, it appears that costs
due to settlement failure are assigned to parties that
have incentives to monitor, should they perceive the
potential for losses to be significant. All that is
required is a commitment by the lender of last resort
to supply net creditor banks the necessary liquidity
to prevent a settlement failure from becoming a
systemic failure, and the proper collateralization to
insure that lending can proceed.
Recently, the Board of Governors has established a
risk reduction policy for large dollar transfer networks. Although this policy relies largely on nonprice risk control measures, the Board made it clear
that it intends as a matter of long-term policy “to
reduce further the volume and incidence of daylight
overdrafts and other uses of intraday credit.“40 A s
experience is gained with risk control policies, it may
be desirable in the future to consider measures that
provide additional economic incentives for banks to
take account of risks they create.

39

In addition, networks
50 Fed. Reg. 21120 (1985).
requesting net settlement services from the Fed will be
required to institute bilateral net credit limits.

26

40

Ibid., p. 21121.

ECONOMIC REVIEW, WAY/JUNE 1985

References
Arrow, Kenneth J. “The Organization of Economic
Activity: Issues Pertinent to the Choice of Market
versus Nonmarket Allocation.” In U. S. Congress,
Joint Economic Committee, The Analysis and
Evaluation of Public Expenditures: The PPB
System, vol. 1, pp. 47-64. Washington, D. C.:
Government Printing Office, 1969.
Association of Reserve City Bankers. Risks in the
Electronic Payments Systems. Washington, D. C. :
ARCB, 1983.
The Final Report of the Risk Control Task
Force. Washington, D. C.: ARCB, 1984.
Baxter, William F. “Bank Interchange of Transactional Paper : Legal and Economic Perspectives.”
Journal of Law and Economics 26 (October 1983) :
541-88
Board of Governors of the Federal Reserve System.
“Statement on Priced Services.” Federal Reserve
Bulletin 70 (September 1984) : 707-13.
Calabresi, Guido. The Costs of Accidents. New Haven:
Yale University Press, 1970.
Chicago Clearing House Association. Daylight Overdrafts. Chicago: CCHA, 1984.
Cook, Timothy Q., and Thomas A. Lawler. “The Behavior of the Spread between Treasury Bill Rates
and Private Money Market Rates since 1978.”
Federal Reserve Bank of Richmond, Economic
Review 69 (November/December 1983) : 3-15.
Hey, John D. Uncertainty in Microeconomics. New
York: New York University Press, 1979.

Humphrey, David B. The U.S. Payments System:
Costs, Pricing, Competition and Risk. Monograph
Series in Finance and Economics, no. 1984-1/2.
New York: Salomon Brothers Center for the Study
of Financial Institutions, Graduate School of Business Administration, New York University, 1984.
Smoot, Richard L. “Billion Dollar Overdrafts : A
Payments Risk Challenge.” Federal Reserve Bank
of Philadelphia, Business Review (January/February 1985), pp. 3-13.
Stevens, E. J. “Risk in Large Dollar Transfer Systems.” Federal Reserve Bank of Cleveland. Economic Review (Fall 1984), pp. 2-16.
U. S. Congress, House of Representatives. Federal
Reserve Competition with the Private Sector in
Check Clearing and Other Services. Thirty-Fifth
Report by the Committee on Government Operations. H. Rpt. 98-676, 98th Cong. 2d Sess. (Washington, D. C.: Government Printing Office, 1984).
U. S. Department of Justice. In re Proposals to Reduce
Risk on Large-Dollar Transfer Systems. Docket
No. R-0515. Comments of the United States Department of Justice. November 16, 1984.
Van Ness, Gardiner B., III. “An Analysis of Broker
Loans.” Thesis, Stonier Graduate School of Banking, Rutgers University, 1975.
Walter, John R. “The Federal Reserve as a Provider
of Payments Services.” Federal Reserve Bank of
Richmond, 1984. (Typewritten.)

FEDERAL RESERVE BANK OF RICHMOND

27