The Virtues of Big Circles
         R. Bass

Consider the problem of multi-layered enforcement arrangements -- where A is responsible for following a rule, B is responsible for seeing that he does so, C for making sure that B does her job, and so on. B is the enforcer for A's compliance, C is the meta-enforcer -- that is, enforces B's compliance with the rule requiring her to enforce A's compliance, and there seems to be no limit in principle to how many layers of enforcement and meta-enforcement there may be. The worry seems to be that as long as you've got a finite population -- i.e., always -- sooner or later you will end up with a complete circle of enforcers that loops back so that A ends up being responsible indirectly for enforcing A's own compliance with the rule. And, if that's so, then there's really no enforcement at all.

Technically, that's not quite right. The circle of enforcers might not include A. It could be that B enforces A's compliance, that C enforces B's, that D enforces C's and that B enforces D's. But it's right in spirit. You still end up with a circle, even if not necessarily one that includes A.

There's another worry, too. Suppose that circular enforcement does not really amount to enforcement at all. Then, if there must be enforcement -- and it's hard to imagine a social order without it -- the only alternative would appear to be some kind of non-circular enforcement. There would have to be some ultimate enforcer who was not herself subject to enforcement, something like a Hobbesian Sovereign with absolute power. There would be no opposing force nor any feasible combination of forces able to resist her, and the Sovereign would be bound only by her own judgment.

So, must we conclude that the alternatives are only either no enforcement of social rules or else dictatorship? Or, in other words, is circular enforcement really a problem? I think it is (in general) only if you make a further implausible assumption. If you assume that no one ever complies with a rule or enforces one or enforces a meta-rule (etc.) unless enforcement is brought to bear upon him, then the problem emerges in full force. However, if compliance or enforcement ever occur for any reason other than because the complier or enforcer is himself subject to enforcement, the problem dissolves.

An enforcer may just think that it's a good idea to enforce the rule or that it's a good idea some of the time or this time or that it's a way of venting frustration from some other source or .... If he does, he'll enforce the rule, at least on that occasion, without having to be forced to by a meta-enforcer. When he does, the next person in the chain will have to do his job, and so on. The circle of enforcement will be broken by the action of someone in the circle who enforces without needing enforcement himself to do that.

There's a further interesting point. The probability that any one enforcer will do his job without being forced to may be quite small, provided the circle of enforcement is large enough. If I consider breaking a rule in the hope that my non-compliance will not attract any enforcement measures against me, then I have to consider the probability that anyone else in the circle will engage in enforcement for some reason other than because he faced enforcement himself if he did not. A very small chance that any single enforcer will may imply a very large chance that some enforcer or other will.

Suppose there are a hundred enforcers before the circle loops back to me and that each has only a 1% chance of enforcing the particular rule relevant to my non-compliance without himself being forced to. Then, there is better than a 63% chance that some enforcer or other will set the process in motion. As a rational decision-maker, I should then -- at least! -- heavily discount any gains I expect from breaking the rule. Of course, larger circles make it even more likely that enforcement against me will be triggered.

In fact, things are more complicated than this in several ways. One is that there isn't just a single circle impinging upon the actions of a given agent. Each is part of multiple circles, tangled and interlocking in various ways, and the agent will only rarely have a clear idea even of how many enforcers are involved that are relevant to a given action or whether that circle of enforcers loops back upon him at all. Second, the same kinds of considerations apply to the other agents, the enforcers and meta-enforcers, as well. They, too, are enmeshed in networks of enforcement and meta-enforcement and face a high probability of sanctions if they don't do their jobs. So, they, too, are well-advised to comply without waiting upon actual enforcement -- which, in turn, increases the probability that actual enforcement will be brought to bear upon any others who do not comply. A third relevant fact, relevant both to the likelihood of spontaneous compliance and of unenforced enforcement is that there is no reason to suppose that all compliance is onerous to the one who complies. Only if it were would there be reason to suppose that only enforcement (discounted by its probability) keeps people in line.

So, with relatively non-onerous rules, what you should expect is that there will be substantial compliance at all levels with very little actual enforcement.¹ There is, in any case, no reason to expect a regime of rules and enforcement to break down or to depend on some highest-level enforcer against whom no enforcement is feasible, such as a Hobbesian Sovereign.

Comments? I'd love to hear.