Let me just continue with Friedman for a minute here. (I really should read his book in full.) On the next page he goes:
… law in the United States and similar systems requires a high standard of proof (“beyond a reasonable doubt”) in a criminal case but only a low standard (“preponderance of the evidence”) in a civil case. Why? The answer cannot simply be that we are more careful with criminal convictions because the penalties are bigger. A damage judgment of a million dollars, after all, is a considerably more severe punishment for most of us than a week in jail.
Economics suggests a simple explanation. The typical result of losing a lawsuit is a cash payment from the defendant to the plaintiff. The result of being convicted of a crime may well be imprisonment or execution.
A high error rate in civil cases means that sometimes I lose a case I should have won and pay you some money and sometimes you lose a case you should have won and pay me some money. On average, the punishment itself imposes no net cost; it is simply a transfer.
A high error rate in criminal cases means that sometimes I get hanged for a murder I didn’t commit and sometimes you get hanged for a murder you didn’t commit. In the criminal case, unlike the civil case, one man’s loss is not another man’s gain.
Punishment is mostly net cost rather than transfer, so it makes sense to be a good deal more careful about imposing it. (Law’s Order, 9-10)
If one is so staunchly utilitarian as Friedman is trying to be, then he should really go hardcore. It is not true that a criminal punishment has no benefits. For one, it helps to reinforce the deterrent stimulus to the rest of the populace.
Let it be that people think the acquitted person is innocent, though in fact, unbeknownst to all (including the jury and judge), he is guilty; or think the guilty man got what was coming to him, though in truth someone else committed the crime. Similarly, everyone is equally strongly convinced that the civil case was decided correctly and did not “lose faith in the system” because of an obviously perverse outcome.
In such a case, it is not true that an error in the criminal case is worse than an error in a civil case from the utilitarian point of view. A man who got hanged for a murder he did not commit (but such that everyone is convinced of his tremendous guilt) was sacrificed for the sake of society, and presumably, if the judge’s decision was well-calculated, society benefited more in terms of criminal acts deterred than the criminal was harmed. Therefore, overall, errors in criminal trials are no more net costs than errors in civil trials.