David R. Henderson  

Multiple Bandits are Roving Bandits

Eighty years later and the res... A Private Non-Subsidized U.S. ...

What I learned from my local anti-tax activism


One of the late Mancur Olson's most-valuable insights was his distinction between roving bandits and stationary bandits. His basic idea was as follows.

Consider an anarchist society in which there is no defense from bandits. (Olson argued that because of the public good/free rider problem, people in a large anarchist society would be unable to defend themselves from a powerful bandit. We can argue about that, but that point doesn't matter for where I want to take this argument.) Then there can be many bands of roving bandits who try to steal from people. Each gang has little or no incentive not to kill the goose that laid the golden egg because if they refrain from stealing "too much," another gang will take it. This is essentially the "tragedy of the commons" problem.

But now, says Olson, imagine that one tyrant manages to take over an area and provides defense from roving bands of bandits. He then becomes a stationary bandit. He'll steal a lot for himself but he now worries about the incentives of people in his territory to produce. He will take the maximum he can, but will think long run. He knows that as he raises tax rates, the incentive to produce falls and, therefore, the amount produced falls. He will maximize taxes, which equal the tax rate times the amount of output taxed. In other words, he will reach the top of the Laffer Curve. (I'm surprised, by the way, that Olson, who must have been familiar with the Laffer Curve, doesn't mention it on p. 469, the place where it would have been appropriate to reference it. I wouldn't be surprised if he actually did mention it but a referee or editor at the American Political Science Review insisted that the reference be cut. One cannot understand American academia without understanding the prejudices, snobbishness, and parochialism that are part and parcel of it.)

In short, while both the roving bandits and the stationary bandits steal, there will be more theft when there are roving bandits.

Where am I going with this? As regular readers of this blog know, I am politically active at a local level, fighting off seriatim the various attempts to increase taxes on the residents of Pacific Grove and of Monterey County. My allies and I have been quite successful. (See here, here, and here. For my long story about my first bit of successful anti-tax activism, see here.) The latest tax increase that we successfully fought off was a proposal for a new bond issue, financed by increased taxes on property owners. The money would have gone to Pacific Grove schools to buy computers, iPads for students, etc.

After we won, I happened to check the web site of the pro-tax side. The site is down now, for obvious reasons, but at the bottom of this link, you can see the list I saw on the other web site. What I noticed was interesting. Whereas I had expected to see the "usual suspects," that is, the various prominent local people who had lined up in favor of tax increases in the past, I saw some of them but very few. Almost all the supporters, to the extent I recognized the names, were teachers, school board members, and other employees of the school district. Noticeably absent were the mayor, members of the city council, and policemen. The one big exception was a retired police chief, but he is married to a teacher.

That was when the Eureka moment hit. It made sense that the "usual suspects" weren't listed. They want to tax Pacific Grove residents for their favored causes, one of the main ones being the huge pensions that the city council (illegally, it turns out) granted to police early in the last decade. If the police and firemen support this tax increase, that will make it harder to get a tax increase further down the road for those pensions.

But the people pushing for the tax increase for schools have different goals. They want to tax Pacific Grove residents for those goals. Thus we have multiple bandits. Each band of bandits wants to tax for its goals. But because each band does not have to worry (much) about how much is left for the other band to tax, each band goes for its favored tax for its favored cause. Multiple bandits are like roving bandits.


Jon Klick has e-mailed me a link to his article, co-authored with Francesco Parisi, in which they make the same point. They didn't, though, link it with Olson's work.

Daniel Kuehn, in the comments below, makes a good point. Yes, referenda do rein in the proclivities of those multiple bandits. But voting, informing, and getting out the vote are costly, and the people doing that, like my allies and me, are saving ourselves only our pro-rata share of the savings from preventing the tax. So we are creating a public good and, as Daniel and other economists know, there is typically underinvestment in public goods; those pushing the tax typically are getting a disproportionate share of the proceeds.

Comments and Sharing

COMMENTS (18 to date)
LD Bottorff writes:

This is an interesting way to look at the problem.
However, we don't need to use the negative term 'bandit' to see the problem here. We should be very skeptical of 'special' tax districts in general. If a local government is supposed to provide funding for all public functions, then they need to do it within the constraints that the rest of us do. We all have to balance what we want against what we have available. Government should do the same.
We all need to be able to prioritize.

Daniel Kuehn writes:

Isn't this a lot of the idea of ballot initiatives? They may support the measure, but the tax can only be imposed if a majority vote for it. Democracy makes the "bandit" singular (not to mention it decentralizes the bandit himself - taxes for computers for kids isn't exactly paying tribute to a gangster! - but that another discussion).

Bostonian writes:

In Massachusetts there must be an "override" vote to increase local property taxes by more than 2.5% (due to Proposition 2 1/2). In my town overrides usually fail, except when there is state money at risk. In particular, a system where states pay for most of the cost of building a new school building if the town chips in the rest encourages local tax increases. If towns had to pay for new school building themselves, they would be less lavish.

JLV writes:

in re: Laffer Curves, a few years ago I tried to convince colleagues to start calling it the Ibn Khaldun curve. Didn't take off, unfortunately.

Jim Caton writes:

Daniel: What do you mean by "democracy makes the 'bandit' singular"? The vehicle of choice may be singular, but bandits are numerous.

Daniel Kuehn writes:

Jim -
I don't understand your point. The mere fact that I am a constituency with an opinion persuading the public does not make me a bandit. I'm a bandit if I can extract whatever I want whenever I want. That shows how strained the "bandit" model is in the first place, but of course no model is ever intended to be perfectly accurate - it's intended to illustrate an important point or mechanism. But no, I think you go very much off the rails when you confuse constituencies within a society with bandits.

David -
Right, clearly other issues get introduced by public initiatives. But a multiplicity of constituencies does not mean a multiplicity of bandits unless they actually each have some agency. Whether certain institutions are efficient or inefficient is another important question.

Daniel Kuehn writes:

In some ways, federalism itself introduces a multiple bandits problem because you have at least three taxing authorities on any one citizen in most cases. What's interesting is that there has been some natural coordination between them to deal with this, with different levels taxing different things. I don't know how many people are under four taxing authorities or more. It would be interesting to test whether they pay more taxes on average.

Another interesting question is exactly who this bandit is (which I've alluded to above). We all know the benevolent social planner that maximizes utility is naive (although in most cases probably not "naive" so much as "oversimplified"). But by the same token the bandit that maximizes tax revenue is also naive. Real polities are some sort of linear combination of the two. What that combination is is largely determined by institutions and it probably matters more than most other elements of the problem.

(of course even that linear combination is an oversimplification... but it's a place to start).

Daniel Kuehn writes:

American federalism, of course.

Jim Caton writes:


I'm not sure I've gone off my rails since I was simply asking a question and restating the premise of the post.

That shows how strained the "bandit" model is in the first place, but of course no model is ever intended to be perfectly accurate - it's intended to illustrate an important point or mechanism.
Which is why David's application of it does carry some legitimacy. Reality always contains restraints not reflected in our models. The question is "how applicable is the model to the scenario in question"? Are you suggesting the answer to that question is "not at all"?
Daniel Kuehn writes:

I meant your sentence, not the question. I had the same criticism of David, why wouldn't I think you're going off the rails when you restate the conflation of constituencies with bandits?

I suppose a multiple bandits model is better than a randomly selected model, so no, not "not at all". But surely what we'd want to ask is whether it's more applicable than an equally parsimonious alternative. And my position on THAT question is that it is not.

Jim Caton writes:

Thanks for the clarity.

Your point about agency is significant to the conversation. Agency exists by association with the school board/system. The ability of members to act as bandits is highly constrained and relies largely on their ability to persuade voters, especially leaders in other agencies (Fire, Police, City Council, etc...). Models of political competition probably better apply to this problem outright, but I think there is value in attempting to nuance the bandit model to fit this scenario.

Motoko writes:

Bandits cannot extract "whatever they want whenever they want". If a bandit tries to steal my stuff, but fails (maybe because I repel him), ze is still a bandit.

Pajser writes:

Olson is charmy, but the claim that small groups are voluntarily formed and big groups aren't isn't convincing. The leaders can occur in group of three, even two men. In small groups, leader may eat the best parts of prey alone. The group can survive if leader is moderately violent and unjust; members may still want to stay in association, because it is safer or something else. Such groups can grow. There is no need to assume hypothetical non-state condition and then introduce two kinds of bandits to explain formation of state.

MikeP writes:

Whatever the merits of modeling the activity as bandits or not, I think a smart elucidation of this observation would make some pretty damning campaign literature.

David R. Henderson writes:

@Daniel Kuehn and Jim Caton,
My reasoning is ceteris paribus. Of course, voters get to vote on these things and, even better, there are supermajority requirements. But hold constant the fact that voters get to vote, and now have multiple parties, rather than one, who can get their tax-increase proposals on the ballot. Then all my reasoning follows.

James writes:


No one needs to assume what you suggest. The earliest evidence of organisms resembling modern humans precedes the earliest evidence of institutions resembling modern states.

If you want to call it a state when just one man starts to bully his neighbor, then you don't really have an argument against Olson. You've only managed to change the subject.

Pajser writes:

     James, I am not sure that I understand you. I agree that no one needs to assume hypothetical non-state condition and two kinds of bandits. But it is what Olson assumed. I do not know what to say about "The earliest evidence of organisms resembling modern humans precedes the earliest evidence of institutions resembling modern states." I do not think humanity is older than state. I think that first humans (and possibly their direct ancestors, Australopithecus) lived in tribes and that these tribes already had all essential elements of the state. Even chimp tribes do. Difference is only in size and complexity.

     Olson's claim that tribes cannot grow, but that there must be some intermediate stateless period seems unsubstantiated. He bases this claim on another claim that small groups can organize collective actions, and large groups cannot. I think I explained how tribe can develop institutions like government and tax without losing its voluntary character; leader starts taxing other members of the group and members still stay in group. They didn't lost their ability to leave the group just because tax is introduced.

John Becker writes:

Hans Hoppe has a book arguing that monarchy represented private government while democracy represented public government. The incentive of private government was to take the long view and maximize tax revenue over generations while public government has an incredibly short time horizon and just tries to maximize what they can take right now. Since democratic politicians are in and out of office, they are like roving bandits in your story while monarchs are like stationary bandits.

Comments for this entry have been closed
Return to top