FRBSF Economic Letter
2002-27; September 20, 2002
Why Do Americans Still Write Checks?
In Europe and other industrialized parts of the world, electronic payment
mechanisms have largely replaced checks. But in the U.S., paper checks
are still very common, accounting for more than 60% of retail payments.
Why don't Americans use electronic payments mechanisms more? One possible
answer is that U.S. consumers simply prefer checks to the available electronic
payments technologies. In that case, there is no efficiency loss from
the continued dominance of paper checks.
An alternative answer stems from what economists call "network externalities."
Electronic payments form a network in which consumers and banks exchange
payments with other consumers and banks that also accept the technology.
Therefore, electronic payments systems are useful only if other entities
also use them. So this network externality is a little like the chicken-and-the-egg
problem. Specifically, consumers would find it preferable to use electronic
payments if many other consumers do, but since a critical mass has not
yet adopted the technology, consumers largely do not use it. If network
externalities explain why U.S. consumers do not use electronic payments
technologies, then there are important implications, insofar as such technologies
increase the efficiency of the payments system and lower transaction costs.
For instance, this explanation implies that encouraging people and banks
to use electronic payments technologies could make others better off,
and it also implies that electronic payments products may be underused.
The U.S. has several electronic payment technologies, such as point-of-sale,
debit cards, and credit cards. In this Economic Letter I focus
on the extent of network externalities for the electronic payment technology
known as the automated clearinghouse (ACH). ACH is a technology that was
developed by the Federal Reserve and is typically used for small recurring
payments between consumers and businesses, such as payroll deposits or
mortgage deductions. The results of the analysis suggest that network
externalities do play a role in the adoption of ACH.
The evolution of ACH
The Federal Reserve has been involved in ACH technology since the early
1970s, and currently has about over 85% of the market share of ACH transactions;
other ACH processors include Visa, Electronic Payments Network, and the
American Clearing House. ACH forms a network because both the originating
bank and the receiving bank or its designated correspondents must accept
ACH to complete the transaction. Presumably, the underlying originating
and receiving customers also must be willing to accept ACH for their payments.
Figure 1 shows that, although the number of ACH transactions processed
through the Fed has been growing since 1990, it is still much smaller
than the number of check transactions. These data are bolstered by estimates
for private providers, which process twice as many check transactions
as the Fed but only about a third as many ACH transactions.
If there is a fixed cost for consumers and banks to adopt ACH, then when
one consumer or bank adopts ACH, other ACH adopters are better off, since
they can exchange payments with one more party. Therefore, consumers and
banks will be more likely to adopt ACH the more other parties do. Moreover,
because ACH usage is technologically intensive, additional users may improve
"user friendliness," which may lower the costs of usage to other
consumers. Both effects lead to an externality in which one user's increased
usage of ACH makes other users better off.
Testing for network effects
To understand the scope of any network externalities for ACH, one must
try to predict how much the user's decision to adopt ACH depends on the
number of other users who have adopted ACH. The ideal way to predict this
would be to fix two sets of users, A and B, require B to adopt or unadopt
ACH, and then observe the extent to which A adopts ACH as a function of
B's decisions. In general, however, ACH adoption decisions are made simultaneously
by all individuals in a network, so we cannot observe situations where
we choose B's usage decisions and then observe A's usage decisions. This
makes it challenging to devise methods to infer the magnitudes of the
In Gowrisankaran and Stavins (2002), several methods are proposed to
test for network externalities in the ACH market. All the methods postulate
that the network for transactions is geographically localized. Their simplest
idea is to examine whether there is geographic clustering in ACH adoption
and usage. The rationale for this test is that, if there are network externalities,
then one bank will be more likely to adopt ACH if other banks in its network
have adopted it, and with localized transactions, this will yield geographic
The study finds significant evidence that ACH adoption decisions are
clustered. But a pattern of clustering itself need not imply network externalities.
For instance, clustering in some parts of the country also may reflect
the fact that those areas are more willing to accept technological products.
In that case, if users B in an area that is less willing to accept technological
products were required to use ACH, this would not make users A in that
area more likely to adopt ACH. While the above study uses data that vary
over time, which partly address this issue, it is also useful to consider
other methods to test for network externalities.
using market concentration
Another method to test for network externalities is to examine ACH adoption
decisions as a function of the local market concentration in the banking
industry. Consider the highest level of concentration, a monopoly bank:
it is not affected by the chicken-and-egg problem inherent in a network
because it is not waiting for other banks to adopt ACH before it chooses
to adopt it itself. This implies that, all else equal, a network with
a monopoly bank is more likely to adopt ACH than is the same network with
two or more banks. Given this description of a monopoly bank's behavior,
economists say that it can internalize the network externality. This story
may explain why banks in Europe, which has a very concentrated market
structure, have been early adopters of electronic payments technologies.
More generally, we would expect that networks with higher industry concentration
will have more ACH adoption, holding fixed other characteristics of the
network, such as the size of the population and of the banks.
To test for concentration effects, Gowrisankaran and Stavins define a
market as a metropolitan statistical area (MSA) and use the Herfindahl
index measure of industry concentration. In the Herfindahl index, a value
of 1 corresponds to a monopoly, 0.5 corresponds to a market with two equal-sized
firms, 0.33 to an industry with three equal-sized firms, etc. Figure 2
shows the deposit-weighted fraction of banks adopting ACH across several
different markets as a function of the Herfindahl index and population
size. The figure shows that, within any population range, more concentrated
markets—that is, those with a higher Herfindahl index—are more likely
to adopt ACH, which is consistent with the presence of network externalities.
Gowrisankaran and Stavins (2002) use more complex econometric methods
to recover the magnitudes of the network externalities implied by this
pattern, which they find to be quite large. From the figure, one also
can see that markets with larger populations are more likely to adopt
ACH. The reason for this is that large markets have large firms, and large
firms can spread the fixed costs of adoption over more customers.
and using small branch variation
A third method of testing for network externalities is to try to find
a set of banks B whose adoption decisions are made independently of the
decisions of banks A in the network but affect the adoption decisions
for banks A. While it is difficult in general to find such a set, the
fact that banks have multiple branches implicitly creates such a set.
To see this, consider a small branch of a large bank that is located in
a town isolated from the bank headquarters. The large bank, which must
adopt ACH for all of its branches together, is not likely to adopt ACH
in response to conditions in the town where its small branch is located.
Thus, a local bank A that is based in the town will think of the adoption
decisions of the branches of the big bank B as unaffected by its own decisions
of whether to adopt ACH.
To implement this test, I follow Gowrisankaran and Stavins (2002) and
find a set of small markets that contain exactly one local bank A and
one or more branches of nonlocal banks B, where the branches are sufficiently
small relative to the size of bank B not to affect bank B's decisions.
The results indicate that if fewer than half the nonlocal banks adopt
ACH, the fraction of local banks that also adopt ACH is about 21%; if
more than half the nonlocal banks adopt ACH, the fraction of local banks
that also adopt ACH rises to 67%. Consistent with the story that a monopoly
network is able to internalize the network externality, 58% of local banks
that are alone in the market adopt ACH. Gowrisankaran and Stavins (2002)
derive formal estimates of the magnitudes of the network externalities
from this sample and use them to compute the expected probability of adoption
if there were one small bank versus two or more small banks in the market.
They find that the network with one small bank will adopt ACH 62% of the
time, while the adoption probability will be at most 46% with two small
banks, and 43% with three small banks. This again suggests that the magnitude
of the network externalities is substantial.
Several different tests provide evidence of substantial network externalities
in the market for ACH. This implies that part of the reason for the continued
dominance of paper checks is a chicken-and-egg problem, where ACH is not
more widely used because of a lack of a critical mass of users. Most new
payment technologies are inherently networks. Thus, insofar as new payment
technologies can increase the efficiency of the payments system and lower
transactions costs, there is a potential concern that network externalities
may cause banks and consumers to underadopt many of them. The results
suggest that we should consider mechanisms that might facilitate the adoption
of new payment technologies by allowing firms to internalize the network
Gowrisankaran, Gautam, and Joanna Stavins. Forthcoming. "Network
Externalities and Technology Adoption: Lessons from Electronic Payments."