Bryan Caplan
July, 1993
* Outline of a Critique of Tyler Cowen's "Law as a Public Good" *
1. Summary of Cowen's Argument
Many critics of free-market anarchism have argued that
collusion rather than competition would prevail, making anarchism no
more attractive than government. However, up to now very little
attempt has been made to justify this claim. On its face, the idea that
defense services are a natural monopoly is highly implausible, as
David Friedman points out in his _Machinery of Freedom_.
It is for this reason that Tyler Cowen's recent critique of
anarchism is a major contribution to the debate. For in his "Law as a
Public Good: the Economics of Anarchy," Cowen puts forward a
powerful reason to buttress the view that collusion would prevail.
There are certain industries, which we may call network industries,
that have a peculiar feature: competing firms must also cooperate (to
some extent) with their supposed competitors in order to be in
business in the first place. Normally, competitors have no reason to
trade with each other -- why should one supermarket have any
dealings with another? But this is not always the case: in industries
like credit cards, banking, and franchises, competing firms must also
cooperate. The member banks of VISA, for example, all require the
services of the VISA central computer, and all depend upon the
acceptability of VISA cards in general. In banking, one vital aspect of
the business is check-clearing done through clearinghouses: each bank
must be willing to honor its competitors' checks and be willing to
balance accounts via a clearinghouse. The case is similar with
franchises: while individual restaurants don't need to contract with
each other, all of them need to contract with the central McDonald's
corporation, which in turn sets common standards.
Well, so what? Why does it matter that in certain industries,
competitors must also cooperate to a limited extent? The answer is
that this industry structure can make collusion work. The usual
checks on collusion are cheating and new entry. But a network can
punish cheating by expelling non-colluders from the network, thereby
putting cheaters out of business. Similarly, the network could refuse
to admit new entrants, thereby preventing new entrants from getting
into the business. The network short-circuits the market checks on
collusion. The network could also solve other checks on collusion such
as the hold-out problem: recalcitrant members could be punished by
expelling them from the network. (A fourth problem -- which we
may call the "division of spoils problem" -- is that it is difficult for
colluders to decide who gets what cut of the monopoly profits. Large
firms often favor the status quo, small firms want to grow, etc. It does
not seem that network industries would have any special ability to
solve this problem.)
What does all this have to do with the viability of free-market
anarchism? Cowen's answer is that the defense industry is a network
industry. Just like the industries discussed earlier, defense firms
must cooperate with their competitors (to a limited extent) in order to
do their job. In the event of a dispute, competitors must agree to
arbitrate with their rivals in order to preserve peace. (Moreover, all
of the advocartes of anarchism have forcefully argued that any sensible
businessman would do precisely that.) But the very possibility of
peaceful cooperation between competitors indicates that we have a
network industry on our hands; and such a network, immune to the
usual checks against collusion, is likely to suppress competition in the
mutual interests of the members. In particular, since the defense
industry, taken as a whole, has a near-monopoly on force, the entire
society would be in danger should the various firms in the industry
succeed in colluding. The united defense industry could do whatever it
wanted. Anarchy would be transformed in a state of the worst sort.
So much for describing Cowen's theory. It has, I believe, a
number of gaps in it; moreover, these gaps, while not individually
decisive, add up to a fairly strong critique. As a caveat, these points
are somewhat speculative, and I would not commit to them without
further research on network industries in general. Let us explore
each of these points in turn.
2. Competing Networks
First of all, Cowen neglects the possibility of competing
networks. Granted, that collusion within a network can work; but is it
the case that every industry with network properties automatically
gets just one network? One obvious counter-example is the credit
card industry. VISA, Mastercard, Discover, American Express, and
other competing issuers exist. Some of these, like VISA and
Mastercard, are actually just networks for member banks who do the
real work of issuing the credit cards; clearly, the credit card industry
has strong network features. But nevertheless, competition exists.
Now it is quite possible that the intra-network competition that exists
is due to the influence of the antitrust laws; if VISA tried to set common
interest rates and annual fees for all VISA card issuers, it might find
itself in court. But inter-network competition seems to be standard,
uncoerced competition.
You might reply that the competing networks of credit card
issuers don't need to deal with each other, whereas competing networks
of defense firms do. I'm not sure if the premise is correct; I suspect
that VISA and Mastercard accept payments from one another and have
other dealings with each other. But in any case, once we are down to a
fairly small number of firms, it is possible to cooperate without a
mega-network. With a small number of firms, bilateral contracts or
even simple repeat can more than suffice to yield cooperation. Yet this
form of cooperation doesn't seem to lend itself to the attendant collusion
that Cowen discusses, since there would be no monolithic network to
punish non-colluders; there would only be six, eight, ten, or twenty
roughly equal bargainers trying to reach a limited understanding. To
punish non-colluders would be to provide an uncompensated public
good to the industry as a whole.
3. Two (Inconsistent) Systems of Belief About
Private Supply of Public Goods; Self-Interested vs. Altruistic Boycotts
Cowen rightly points out the contradiction between two widely-
held libertarian beliefs: first, that collusion is impossible; second,
that voluntary efforts can solve the public goods problem. Clearly,
since the former is just a species of the latter (collusion is a public
good for all members of an industry) the beliefs are incoherent.
Right?
Well, not necessarily. Let us distinguish between two kinds of
boycotts: self-interested and altruistic. A self-interested boycott
might be a boycott against cheaters or thieves. Once you learn that a
person is a cheat or a thief, you don't need to be coerced to get you to
respect the boycott: it is in your interest. We may contrast this with a
boycott again, say, uncharitable people. Just because someone is
uncharitable doesn't make them worse customers. A boycott like this
would indeed need coercion (or at least strong ideological motivation)
to work.
This distinction shows a problem with one of Cowen's
arguments. As he puts it, "If the network can implement successful
sanctions against outlaws, however, the network can also implement
successful sanctions against potential competitors. If punishing
potential competitors is too costly, punishing outlaws is too costly."
(p.259) However, punishing outlaws is surely a self-interested
boycott: dealing with them is likely to get you or your employees
killed. But punishing collusion is an altruistic boycott: it benefits the
whole industry, not the individual firm doing it (which actually loses
out by refraining from a profitable trade). Since collusion is rarely
the kind of thing motivated by ideology, coercion would be necessary to
make it work. The theory that boycott can squelch open violence and
outlawry without facillitating collusion is therefore consistent.
As an aside, it seems to me that what Cowen refers to as a
"suspension of quality competition" (p.259) is actually an aspect of
quality, since it brings about the peaceful and orderly settlement
procedure that customers want. And it should probably not be thought
of as collusion, since by conforming to common standards the firms
benefit themselves by benefiting their customers, not by hurting them.
It is also not collusive in another sense, since the firms that cheat
would actually hurt their interests. A self-interested boycott is all
that is needed to enforce adherence to common rules of law and
procedures.
4. Networks by Degrees
A closely related problem is that Cowen seems to think that
being a network industry is an all-or-nothing thing. But in reality,
industries have network properties to varying degrees; there are
strong networks and weak networks. In general, the stronger a
network, the easier it will be for it to make not only self-interested
but also altruistic boycotts stick; and the weaker a network, the more
it will have to rely on self-interested boycotts and despair of the use of
altruistic ones.
Let us take a few examples. Credit cards seems to be a great
case of a really strong network industry. It would be impossible for a
single bank to enter the market on its own and try to set up a new
credit card. It would have to negotiate with millions of establishments
to accept its card, build up a customer base, get a central computer to
process claims swiftly, and so on. For this reason, banks must issue
their cards via a network that takes care of these problems for it. And
if a network became displeased with a bank, it could expel it from the
industry by simply expelling it from the network. In sum: the
network could apply a stinging sanction to members who didn't take
part in an altruistic boycott (e.g., to enforce collusion), so it could
probably make such a boycott work.
From personal conversation, I learned that Cowen thinks that
banks are another strong network industry. I am open to the
possibility, but it doesn't seem too likely. Clearing checks is only one
of the many things that banks do; a bank could continue most of its
activities no matter what the network (in this case, the clearinghouse)
threatened. Moreover, this seems like another industry in which
competing networks are quite feasible: there might be twenty regional
clearinghouses that settle their accounts with each other via bilateral
contracts, or just by repeated relationships rather, rather than with a
meta-network. In fact, unless the history books lied to me, the free
market prior to 1913 obstinately refused to produce a national
clearinghouse, so the Federal Reserve Act had to create one by law.
How did clearinghouses handle inter-regional clearing before the Fed?
I don't know, but I guess that they handled it by bilateral contracts.
Now maybe at that time, clearinghouses were local natural monopolies;
but in modern times, especially with new communications and
computer technology, it seems that banks could easily contract with
any willing clearinghouse to represent its interests, regardless of its
location. So the likelihood that banks could organize an altruistic
boycott (e.g., to enforce collusion) seems low.
To take a third instance, consider professsional associations. It
is hard to be an economist without joining the AEA. Yet the AEA hardly
seems to have effectively enforced collusion in, say, hiring -- even
though it would probably have fewer antitrust problems than a for-
profit firm. Why is this? I would suggest that professional
associations are a weak kind of network: if they tried to impose costs on
its members, the consequences of withdrawal would not be very severe
for the withdrawers. Members with tenure, for example, might just
laugh. Most of the benefits that the AEA provides -- a place to meet and
discuss -- could be carried on with some inconvenience outside of the
network. It seems highly unlike that anything more than self-
interested boycotts could work in such a network.
In sum, then, networks vary in strength from strong to weak;
and while all networks can easily apply self-interested boycotts, it
takes a fairly strong network to enforce altruistic boycotts.
5. Factors Tending to Make a Network Strong or Weak
Now let us consider the crucial question: would the defense
industry be a strong or weak network? This is somewhat related to the
question of competing networks, but it is sufficiently different to be
worth discussing on its own.
What factors contribute to the strength or weakness of an
industry's network structure? This another area where I would like to
do more empirical research, but I have some preliminary ideas to
throw out. It will be noticed that typically, the same factors that make
a network weak also make it more likely that there will be no network
at all, but instead a lot of bilateral contracts.
a. Let us call the first factor "the degree of interrelatedness"
for want of a better term. The intuitive idea this that in some
networks, every firm has dealings with every other firm in virtually
every time period. In banking, for example, every bank has checks
going to or from a huge number of other banks. In a professional
society, in contrast, members might only deal with a small fraction of
the membership. It seems likely that the higher the degree of
interrelatedness, the stronger the network.
b. Closely related is the "lumpiness" of inter-firm relations.
If firms deal with each other rarely but their deals are large, bilateral
contracts and the like may work better than a network. In contrast, a
large number of small deals make a network more efficient. We can
think of the same fact in another way. If deals are lumpy, then a
network would involve wasteful transactions costs, since every firm
would be paying to make it possible to deal with every other firm, even
though such deals rarely materialize. In contrast, if deals are less
lumpy, firms waste transactions costs by negotiating a new deal with
each trading partner.
c. Networks that provide services vs. networks that just
coordinate cooperation. The VISA corporation has a giant
supercomputer that processes the hundreds of millions of monthly
VISA transactions at a very low per unit cost. No individual member
could afford/find it profitable to buy such a computer. VISA also
invests in a brand name, common advertising, etc. In short, there is
something more this network than just the members: there are
common resources that the network as such owns, and which it is
uniquely able to supply. In contrast, other networks have few or no
common resources; instead, they just coordinate the cooperation of the
members. A professional association is probably a good example: it
doesn't supply its members with anything significant; it just solves
the coordination problem by announcing a single time and place for
gathering, letting people spread information, and so on. Networks that
provide services are probably much more able to enforce altruistic
boycotts than networks that just coordinate cooperation.
d. The extent to which inter-firm cooperation is based on
consciously-designed terms vs. evolved traditions. The VISA
corporation is an enormously complex, designed institution. It was
consciously worked out to solve certain problems, and appears to
require fairly frequent modifications, modernizations, and so on. In
contrast, some kinds of cooperation are based more on tacit, evolved
understandings -- e.g., don't use violence first, be polite, reciprocate
favors. Networks based on explicit design are probably stronger, since
they are the only cheap remedy for high transactions costs. But if
cooperation is based largely on evolved tradition, then the network is
likely to be weaker; tradition in a sense substitutes for formal (and
costly) negotiation.
e. The extent to which firms have other activities for which the
network is unnecessary. Credit card issuers simply can't do business
unless they are in a network; but members of professional societies,
banks, and other network industries do a lot of non-network things too.
Banks, for example, could still take savings and make loans even if
they broke with their clearinghouse; professionals could still get jobs
even if they didn't attend conventions, and so on. The more non-
network substitute lines of business the firms have, the weaker their
network will be.
f. The final point is not so much a point about the power of a
network, but rather a point about how likely a network is to use
whatever power it has. Some kinds of collusion are easier to enforce
than others: if the product is homogeneous, monitoring is cheap, and so
on. With heterogeneous products and expensive monitoring, collusion
may not work even if the power to enforce it exists. (See section 3 for
why this factor preventing collusion wouldn't similarly prevent all
order.) For example, malls forms a kind of network: no store can
trade in a mall unless they have the permission of the mall owners.
Now since the cross-elasiticities of demand for the goods sold in
different stores are probably significant, it is clear that in theory all
of the stores could make money money if they colluded (a little bit --
otherwise people would go to a different mall); and since they are in a
common network, this would be feasible. But to my knowledge, no mall
does this. Presumably, goods are too heterogeneous (clothes substitute
for each other, but there is a huge variety of types of clothes) and
monitoring costs are too high (prices need to go up and down, it would
take a lot of watchdogs, etc.).
Perhaps there are other factors, but I can't think of any right
now.
6. The Defense Industry: Strong or Weak Network?
Given this rough outline of factors tending to increase and
decrease the power of a network, where does the defense industry wind
up? I think that it winds up on the low end in almost every case.
a. Degree of interrelatedness. Most crime is local; and most
international crime consists in the drug trade and other activities that
would presumably be legal in an anarchist society. So it is likely that
each defense firm will have almost all of its dealings with other local
firms. L.A. might have, say, 20 firms, each of which would have to
arbitrate many conflicts with each other; but they would probably
have very few conflicts to work out with Boston firms, since very few
criminals travel from Boston-L.A. to commit crimes (or vice versa).
Cowen himself concedes this point without admitting its full
significance: "international criminal and legal relations are a very
small part of the activities of, say, the American and Canadian
governments. The benefits of forming a world government to more
efficiently deal with Americans who murder Canadians are relatively
small." (p.263) International crime is a minor problem because
inter-local crime in general is a minor problem. If most crime is
local, then networks (if they exist) will probably be local too, so at
least a national monopoly wouldn't arise. Moreover, free entry would
be harder to deter if whatever collusion existed were local, since the
possibility of massive retaliation would not be present. It would not be
the System vs. a lone rebel firm, but a conflict between roughly equal
networks who would prefer live-and-let-live to fighting.
b. Lumpiness. Again, this seems to hold. Once you consider
only crimes with victims, you will be surprised to find out how few
crimes there actually are in each locality. Moreover (and this is
perhaps an aspect of lumpiness), each alleged crime is likely to be
different from every other: particular circumstances of time, place,
evidence, witnesses, and so on must be considered to reach the outcome.
It is unclear why a network's transactions costs would be lower than if
two firms just worked it out for themselves. In contrast, the
transactions costs of highly standardized credit card deals can be
reduced a lot by a network, since they can all be treated alike. The
VISA central computer handles millions of transactions, each of them
nearly identical; it average transactions costs are much lower than if
each firm did it. But it is difficult to see how a central trial system
would resolve cases at lower costs. (Maybe if Landes&Posner were
right that providing efficient rules of law is a pure public good would
this be the case. See my thesis "The Economics of Non-State Legal
Systems," sections 3.1-3.8 for my doubts.)
c. Would defense networks provide services or just coordinate
cooperation? Since the main thing that the network is supposed to do is
create orderly channels to resolve disputes between clients of different
firms, it seems like the defense network's chief business is the second.
Cowen points out that criminal files, extradition rights, and the like
might be provided by the network. Quite possibly, but it still seems
like the defense network mainly tries to reduce the transactions costs
that firms would incur if they all made bilateral contracts with each
other.
d. Would industry standards be designed or evolved? To a large
extent, the latter. We have a huge body of customary and common law,
business practice, and so on; we also have widespread adherence to
some simple norms of conduct. So I think that the standards for inter-
firm relations would be basically set by our traditional practices: not
starting conflicts, compensating the injured, fair trials, and so on.
This is just a big parts of our cultural background and no one (in their
private role, anyway) really wants to re-negotitate it. So the
network's contribution to reducing transactions costs wouldn't be that
big. There might be some designed aspects that would require formal
(and high t.c.) negotiations, but we may reasonably doubt thir extent.
It seems likely that designed innovations would be largely intra-firm.
e. Extent of other activities. Defense firms would do a lot
besides work out trials of accused criminals. They would patrol areas,
guard buildings, walk people home, check out false alarms, and supply
a long list of other defensive services that people would want. This
bodes well for anarchy for two reasons. First, the power of the
network is less when firms have easy substitutes lines of business.
Second and more importantly, even if Cowen were entirely right about
the collusive tendencies of the arbitration side of business, that would
be just a tiny sub-market in the broader defense industry. The
weapons would not be monopolized, because only one part of the defense
industry would tend to monopoly/cartelization. And if that sub-market
started to act like a state, people would still have many defensive
alternatives.
f. Heterogeneity and monitoring costs. To begin with, is clearly
easier to enforce rules against violence and stonewalling than it is to
enforce collusion; the former would be a self-interested boycott, the
latter an altruistic one. Violence is more obvious than non-
collusion. And given differences in location, quality, market niche, and
so on, heterogeneity could be a big carrier to collusion even if the
general Cowen picture is correct.
7. Ideology: Deus ex Machina?
Cowen argues that to use libertarian ideology to prevent his
scenario is just a deux ex machina. Perhaps. But suppose that it were
argued that ideology is also necessary sustain government. Many social
theorists have said as much (Hume, la Boetie, Hayek, Mises, and lots
more). In this case positing libertarian ideology as one check on abuse
of power becomes more reasonable.
As I suggested in conversation with Cowen, people seem to
behave "altruistically" ("ideologically" might be a more neutral word)
in one sphere of life but not the rest. In their role as private
individuals, people give huge amounts of money to charity. But very
few businessmen run their businesses like charities, even if they are
charitable people. (Even there, though, most people wouldn't murder
for a living, even if the pay were good.) Perhaps this is a way to solve
Cowen's paradox, that if private supply of public goods works, then
collusion works, and if collusion doesn't work, then private supply of
public goods doesn't work either. If people are (somewhat) ideological
qua private individuals, but profit-maximizing qua businesspeople,
then both good results (non-collusion and private supply of public
goods) can spring from the same group of people.
Another related observation: While in their business roles
people are probably less charitable than in their private roles, most
businesspeople still feel somewhat constrained by morality. Most
wouldn't murder or steal to increase profits. In a way, the business
role tends to relax some, but not all, of the moral constraints that they
feel as individuals. (Not that this is a bad thing -- I agree with Milton
Friedman that managers should be charitable with their own money,
not their stockholders'.) Compare this to the governmental role: this
seems to relax almost all of the moral constraints that people feel. It
is a truism that the gentlest people will kill for their governments;
but how many would kill for their employers? So perhaps this is
another reason (and not just a deus ex machina) to think that
ideological constraints on abuse of power would work better in an
anarchist society.
8. My Picture of Anarchy
Let me outline my picture of the defense industry in an
anarchist society, to contrast it with Cowen's. There would probably
be a middling number (10? 20? 30?) of weak networks; the networks
would be bound to each other by bilateral contracts. Basic standards
like adherence to peaceful and orderly procedures, willingness to
negotiate, fair trials, and so on would be enforced by self-interested
boycotts. Collusive standards, in contrast, would probably be too hard
to enforce for all of the reasons suggested. It also seems likely that
local firms might have more precise and detailed contracts with each
other, since most crime is local in nature.
The incentive to form networks above the local level would
come from the desire for extradition and the like. This is a pretty
minor benefit, but probably worth pursuing. (After all, ostracizing
dangerous criminals is another sort of self-interested boycott.) Peter
Kurrild-Klitgaard suggested to me that there would be many
overlapping jurisdictions, and I am inclined to agree with him.
The worst that could happen is that the criminal arbitration
side of business would become a collusive network; but this would still
leave open competition and protection from well-armed firms in other
lines of business that have no need to network with each other (as I
suggested: armed guards, patrols, etc.). But that is not too likely; not
only are internally collusive but externally competing networks
possible; but for the reasons stated above, the networks are likely to
be too weak to enforce internal collusion either.
Admittedly, this is a speculative picture; but so is Cowen's.
Precise predictions of market structure are not forthcoming, but some
predictions make more sense than others. Despite my view that
Cowen's critique of free-market anarchism is the best yet, I must say
that his prediction is much too pessimistic.