Quis custodiet ipsos custodes?

Powers used to enforce rules may also be used for other purposes—extortion, intimidation of political opponents, facilitating ordinary crimes such as robbery. Who is to enforce the rules upon the enforcers?

The simplest solution is to have no enforcers with special rights or powers. Consider a system of social norms. What prevents me from teaching my classes in jeans is not the fear of being arrested but the expectation that doing so would lower my reputation, social and professional, would change to my disadvantage the behavior of people with whom I interact. Norms are enforced against me by everyone I interact with, by me against everyone I interact with. That describes not only the social norms I face but the norms of neighborly behavior in Shasta County, California, as described by Robert Ellickson in Order Without Law, enforced by both legal and mildly illegal acts against norm-violating neighbors.

A feud system such as traditional Somali law or the legal system of saga-period Iceland also lacks specialized enforcers. The rules are enforced by the threat of the private use of force by anyone against anyone. Some individuals are more formidable or have more allies than others but nobody has any special rights or legal status associated with the job of enforcing rules. Someone who uses his strengths to try to abuse the system risks bringing into existence a stronger coalition to block him.

For a third example, consider England in the eighteenth century. Since any Englishman could prosecute a criminal case, the fact that an offense was approved of by the authorities was no guarantee that it would not be prosecuted. The point was demonstrated when a demonstration in favor of imprisoned radical John Wilkes ended with troops firing into the crowd and killing several people. The Wilkites responded by charging several of the soldiers, the magistrate who had ordered the troops to fire and the other magistrates present with murder.

The king had the power to pardon a convicted felon but doing so in too obviously partisan a way might provoke public outrage. In one notorious case two convicted murderers were pardoned, apparently because their sister’s aristocratic lovers applied political pressure on their behalf (“the mercy of a chaste and pious prince extended cheerfully to a wilful murderer, because that murderer is the brother of a common prostitute”). The Wilkites responded by raising money to fund an appeal of murder, a private criminal case. An appeal was a complex, expensive and difficult proceeding that had gone almost entirely out of use. It had, however, one large advantage:

“If the appellee be found guilty, he shall suffer the same judgement as if he had been convicted by indictment: but with this remarkable difference; that on indictment, which is at the suit of the King, the King may pardon and remit the execution; on an appeal, which is the suit of a private subject, to make an atonement for a private wrong, the King can no more pardon it, than he can remit the damages recovered in an action of battery.”

Blackstone 1884, Bk 4 Ch 23

The appeal failed, as did the earlier criminal prosecutions of the soldiers and magistrates, but like them demonstrated the possibility of using privately prosecuted criminal law against malefactors supported by the government.

A different approach is to have a second layer of enforcers charged with the duty of enforcing the rules on the first layer. In the modern U.S. that may mean a civilian review board to examine charges against police officers. In Islamic law the nazar fil-mazalim, “investigation of complaints,” was a prerogative of the caliphs by which they “or, by delegation, ministers or special officials and later the sultans, heard complaints concerning miscarriage or denial of justice or other unlawful acts of the kadis, difficulties in securing the execution of judgements, wrongs committed by government officials or by powerful individuals …, and similar matters. Very soon formal courts of complaints were set up.” In imperial China, the censorate “had as its primary general purpose the investigating and impeaching of governmental wrongdoing or corruption wherever uncovered.” While this approach may limit the ability of ordinary enforcers to abuse their powers, it raises the risk of abuse by the layer of enforcement above them.

A variant is to use one category of specialized enforcers against another, as when the FBI or state police investigate corruption by local police. A more decentralized version uses tort law to punish abuse of the powers of enforcers of criminal law. Modern law texts describe what police must do in order that the evidence they procure will be admissible. Nineteenth-century textbooks described what police had to do in order not to be sued. That approach is not entirely dead. In a number of prominent cases in recent decades, such as the shooting of Black Panthers Fred Hampton and Mark Clark in Chicago in 1969 and the Steve Jackson case in Texas in 1990, victims of law enforcement abuse sued and collected.

A famous eighteenth-century example of the use of tort law to restrain enforcers of criminal law was Hinkle v. Money, one of the earliest punitive damage cases in English law. Issue 45 of The North Britain, an anonymous anti-government publication, contained an article attacking in strong terms a royal speech.640 The government responded by sending out Kings’ Messengers, the eighteenth-century equivalent of Secret Service agents, with a general warrant authorizing them to arrest any person and seize any papers that they believed were connected with the publication. They arrested forty-nine people. One of them, a journeyman printer, sued, on the grounds that holding him prisoner for six hours while searching his papers on the authority of a warrant that did not name him was illegal. He won the case and collected three hundred pounds from the Secretary of State.

The award was against the Secretary of State but it was paid on his behalf by his government. The awards in the Black Panther and Steve Jackson cases were paid by the governments that the law enforcers found liable worked for. While that pattern holds, tort law provides an incentive to prosecute in the form of a possible damage payment and an incentive for governments to try to control law enforcers, but it does not provide a direct incentive for the enforcers themselves not to misuse their powers.

A further problem with the use of tort law or privately prosecuted criminal law to prosecute misdeeds by enforcers is that the government that employs the enforcers has, and may use, veto power. When it was discovered that AT&T had been providing information to the National Security Agency, arguably in violation of its obligations to its customers, customers sued. Congress responded by altering the law to immunize phone companies from such suits. If the Wilkites had succeeded in convicting a magistrate of murder for instructing troops to fire on a crowd the King could, and very likely would, have pardoned him.

That problem does not exist with a feud system. In Egilsaga, one of my favorite of the Icelandic sagas, Kveldulf’s son Thorolf, a retainer of King Harald, is attacked and killed by the king, worried by Thorolf’s increasing wealth and power. Skallagrim, Kveldulf’s other son, asks the King for wergeld for killing Thorolf and is turned down. Kveldulf and Skallagrim depart for Iceland, then being settled. On the way they stop long enough to intercept and attack a ship carrying two of the king’s nephews, a ship that the King had earlier seized from Thorolf. They kill the nephews and most of the crew and send one survivor back to Harald to tell him who was responsible. That sets off a three-generation-long feud between a family of Icelandic farmers and the royal family of Norway.

East of Kveldulf’s Island was a world his foeman ruled;
West of it a land where men were free.
Death’s the price of living and the Norns are never fooled;
He saw the stolen ship go by and followed it to sea.
He would never look on Iceland but he led proud Harald know
That even kings pay wergeld, though they will not have it so,
He made them brace and bend their backs and row, row, row.

One final approach to controlling enforcers is suggested by David Brin.642 He predicts that improvements in surveillance technology will produce a society where everything that happens in any public place will be recorded and findable, putting enormous potential power into the hands of law enforcement. He argues that the best way to control that power is for transparency to run in both directions, the ability of the police to watch the citizens balanced by the ability of the citizens to watch the police.

Brin’s solution depends on the officials controlling surveillance giving the public access to the resulting data. It also depends on the surveillance covering things the authorities may not want covered; one can easily imagine police, before beating up a suspect, making sure that the room’s video camera was turned off. But recent high-profile cases where law enforcement agents got into serious trouble as a result of bystanders video recording their misdeeds suggest that something along the lines Brin suggested may be occurring. The technology is new but the underlying theory is not. One very ancient approach to controlling the misdeeds of government actors is for the ruler to hold open court at which any of his subjects may approach him to complain of the acts of his officials.

Both the old and the new versions suffer from the same limitation. The police can watch us, we can watch the police. The police can arrest, jail or execute us; we cannot do the same to them.644 Brin’s approach depends on the willingness of rulers to punish misdeeds by enforcers. So did the older version.

H.L. Mencken, in a satirical discussion of the problem of controlling misdeeds by government officials,645 describes two solutions. The German solution was a special court for trying errant officials. It worked because “a Prussian official was trained in ferocity from infancy, and regarded every man arraigned before him, whether a fellow official or not, guilty ipso facto; in fact, any thought of a prisoner’s possible innocence was abhorrent to him as a reflection upon the Polizei, and by inference, upon the Throne, the whole monarchical idea, and God.”

That approach would never work in America, since “even if they had no other sentiment in common, which would be rarely, judge and prisoner would often be fellow Democrats or fellow Republicans, and hence jointly interested in protecting their party against scandal and its members against the loss of their jobs.” Mencken therefore proposes an alternative better suited to American conditions: “… any [American citizen], having looked into the acts of a jobholder and found him delinquent, may punish him instantly and on the spot, and in any manner that seems appropriate and convenient―and that, in case this punishment involves physical damage to the jobholder, the ensuing inquiry by a grand jury or coroner shall confine itself strictly to the question of whether the jobholder deserved what he got.”

Think of it as a modified version of the tort law approach—privately initiated actions to control public enforcers. Or perhaps a modern form of feud law.

Democracy as the Solution—but Only for Pirates

I have so far ignored what many modern people think of as the most important constraint on enforcers—democracy. If the chief of police is corrupt and the mayor refuses to fire him, vote the mayor out. That was, after all, the mechanism that eighteenth-century pirates used to deal with a corrupt or incompetent quartermaster.

For a pirate ship with a crew of a hundred or so, it may have worked. The scale of the polity was small enough so that the individual voter had adequate first-hand information on how good or bad a job the quartermaster was doing. And it was reinforced by a second mechanism—reputation. If the quartermaster pocketed more than his share of the loot and managed to fool or bribe enough of the pirate voters to keep his position, the next time the ship docked at a pirate port it might lose half its crew and have a hard time recruiting replacements.

It works much less well for a city of three million or a country of three hundred million. The average citizen of Chicago or the United States sees directly only a tiny fraction of what the employees of his government, including those enforcing the laws, are doing. In order to decide whether they are doing a competent and honest job he needs to devote a considerable amount of time and energy to gathering information, listening to and evaluating arguments by supporters and critics.

That is a cost, and people are generally willing to bear costs only if they expect benefits from doing so. In a population of three million, the chance that my vote for mayor will change the outcome is tiny, probably less than one in ten thousand. In a population of three hundred million choosing a president, tinier still. That is not much return for an investment of many hours of time and effort. Better to spend my time instead researching what model of car or refrigerator to buy, since in that case my vote is decisive.
It follows that voters in large polities will be rationally ignorant of the information they would need in order to control government actors. As a test of the theory, I have occasionally tried the experiment of asking a room full of college students if they know the name of their congressional representative. Generally about half say that they do. It is hard to keep track of what someone is doing if you do not know his name.

The rational thing for an ordinary voter to do, unless he is a fan of politics as a spectator sport, is to decide which candidate it is in his interest to support, not which will do a better job if elected. That may mean supporting the candidate whom most of the people he interacts with support. It may mean supporting the candidate who produces sound bites that make his supporters feel good. It may even mean supporting the best-looking candidate.