Bryan Caplan  

The Preponderance of the Evidence

PRINT
Interview with Trent McBride, ... The Ingredients and the Recipe...

Years ago I attacked the "reasonable doubt" standard of criminal justice, using the following hypothetical as a starting point:

[I]magine that you are observing a trial for murder. Most of the evidence goes against the accused (who allegedly murdered a 5-year-old girl). The girl's father sits anxiously, awaiting the verdict; he testifies that he witnessed the murder personally, but various other strands of evidence exist which create some doubts.

After witnessing the whole trial, a definite conclusion about the evidence strikes you: the accused is probably guilty, but not guilty beyond a reasonable doubt. And the jury, apparently, felt the same way: they regretfully emerge from deliberations and proclaim the defendant "not guilty."

Suddenly, the enraged father whips out a concealed hand gun and aims it at the smiling acquited defendant. He shoots and kills the defendant. Now the question arises: what, if anything, may be justifiably done to the father?

I go on:
Imagine you were the bailiff for the trial and you saw the father draw his gun; should you shoot and possibly kill the man who probably is not a murderer, in order to save a man who probably is? Only the preponderance rule lets us escape the bizarre conclusion that we should kill the man who is less likely to be a murderer in order to save the man who is more likely to be a murderer.
This still seems right to me; any objections?


Comments and Sharing





COMMENTS (34 to date)
KipEsquire writes:

You are flippantly omitting a "yet" -- "possibly kill the man who probably is not YET a murderer..."

Your hypothetical is utterly unpersuasive.

Andrew writes:

Wait. Shouldn't the question be "should you shoot and possibly kill the man who is about to become a murderer, in order to save a man who probably already is one?" ?

Alex J. writes:

Why doesn't the legal system weight punishments according to confidence in the verdict? The jury is asked two questions: "Guilty?" and "What punishment?" They are asked to answer each question without regard to the other. Various anecdotes suggest that sometimes this separation causes the juries to warp their decisions in various ways.

ryan writes:

Sure, I'll object. I reject the assumption that (absent an anarchic system) anyone has a right to kill a murderer. (If we don't reject this assumption, then, among other things, we must conclude that most crimes committed against other prisoners in jail should not be punished -- we should simply reduce the sentence of the assaulted prisoner and consider the other prisoner to have simply been speeding up the process of punishment.) Therefore, the father is guilty beyond a reasonable doubt, and this conclusion shouldn't trouble us, because even with full certainty of the dead man's guilt, or with a preponderence standard, we'd convict the father.

On the other hand, suppose I support your assumption about the right to kill a murderer. Then remember what you say in that essay -- it's quite reasonable to doubt that he killed a noncriminal who had done him no wrong (that is, the we suspect that he probably killed a murderer and so didn't kill a non-murderer). In fact, it's probably true beyond a reasonable doubt that he at least believed he was killing someone who had murdered his daughter, and so there is no mens rea. Therefore by the standard of reasonable doubt, he is not guilty and we don't need to come to the conclusion you seek to avoid.

richard pointer writes:

In civil law countries a person is often found responsible for a crime depending on the level of guilt accorded his or her actions. That way you get a gradient of guilt rather than a all or nothing result. Too bad we can have a political system set up this way.

Dave writes:

A gradient system doesn't really make things better - it's just a different trade off. Rather than some guilty people going free, you have a portion of innocent people who are punished because they might be guilty but probably aren't.

Tom West writes:

We don't use preponderance of evidence because our society has decided that it is vastly worse to have innocent people unjustly punished than guilty people unjustly free.

The fact that one can come up with scenarios where the expression of that will comes at significant cost doesn't mean a whole lot.

(Or at least any more than examples where failing to legalize torture can be bad, preserving intellectual property rights can be bad, etc., etc.)

Eli writes:

The father has no right to avenge his daughter's death.

If this idea troubles you, consider the murder of someone who has no friends or family. Does it follow that this murder should not be a crime because no living party is harmed? Of course not. The murder is a crime because cooperative institutions have evolved that protect even the friendless.

The father's desire for revenge or recompense is completely irrelevant. The bailiff has a responsibility to shoot the father, and I see no contradiction in him doing so.

Ian F writes:

The fallacy is that you are accepting the father's testimony while omitting the countervailing evidence that convinced an independent jury of the defendant's innocence. There must have been some persuasive evidence to convince them that the defendant most likely did not commit the crime, outweighing the father's eye-witness identification.

Moreover, just because the father disagrees with the ruling does not make him innocent of the murder that he subsequently commits (with multiple witnesses). We have the legal system so that people do not mete out justice on their own, whether out of malice, spite , or even personal senses of justice. To change your hypothosis around only slightly, imagine that the daughter is an unborn fetus and the defendant is a doctor. Even if the father believes that abortion is murder, society deems the doctor "innocent." Killing him to avenge the death of his daughter would be unacceptable in either case. THis is not the Wild West anymore.

More importantly, we have a justice system that leans towards allowing the guilty to go free, rather than the innocent to be jailed. That's partially why we are innocent until "proven" guilty. The father could not "prove" it to his peers, so the defendant went free.

Mark writes:

A few points:

- Yes, your hypothetical presents a bad situation. However, in matters of the justice system (and society in general) there are rarely perfect solutions. The "beyond reasonable doubt" standard will result in more (actually) guilty people going free. Conversely, the "preponderance of evidence" standard will result in more innocent people getting convicted (I shudder to think how many weak cases would be brought to trial if prosecutors knew they only had to meet the preponderance standard). You'll have to do more than present one bad situation that arises as a result of "beyond reasonable doubt" to make your case: you'll have to show that, on balance, a different standard leads to overall better results.
- It's not clear to me that "preponderance of evidence" resolves your dilemma in any case. Under a more relaxed standard, you could still have to render a not guilty verdict if you felt the evidence indicated there was only a 49% chance the defendant committed the crime. He's still much more likely than the father to be a murderer, but the bailiff would still have the duty to shot the father in order to protect the defendant.

ryan writes:

Ian, I don't think leaving out evidence is the issue here. It's not that the jury was persuaded otherwise -- they're almost certain that the dead guy was a murderer. The point is that "reasonable doubt" means that if you're 90% sure the guy is a murderer, you let him go (and so the jury is 90% certain that the father killed the guy who killed his child)

Paul Zrimsek writes:

Does "What on earth are you talking about?" count as an objection?

Matt Peppe writes:

Even if, in this particular case considered alone, it may be preferable to let the father shoot the criminal, it's still better to shoot the father on the whole.
If you don't shoot the father, you've broken the system. More people are now more likely to follow his example and take personal vengeance against those who the law would set free.

SheetWise writes:

"Only the preponderance rule lets us escape the bizarre conclusion that we should kill the man who is less likely to be a murderer in order to save the man who is more likely to be a murderer."

With the assumption that the prosecutor is a more reliable source of information than the defense. A dubious assumption. To be clear -- the prosecutor may have a "Nifong" problem.

If I were the father -- I would be more concerned with the state law -- particularly those on the fully informed jury.

While denying an accused the right to face his accusers may seem severe -- there are many situations where prosecutors will not proceed if they expect the jury has the right to nullify.

Nullification should be embraced along with full information on the role of a jury -- it would allow most of us to honestly "Imagine you were the bailiff ...", and act accordingly.

Constant writes:

Is the proposed rule that it is always right to shoot someone whose probability of guilt is > 50% but otherwise not?

Suppose that I possibly (possibly not) shoot someone whose probability of guilt in a previous murder is only 45%, but then at my trial the evidence against me only comes to 55%. So from the jury's point of view, there is a 55% probability that I killed the other person, but in that case there is still a 45% probability that I killed a guilty man. So from the jury's point of view, I probably did not kill an innocent man. That is, the probability that I shot an innocent man is 55% times 55%, or about 30%.

Am I to be found guilty and punished or not? It depends on what the rule is. There is a 55% probability that I am guilty of violating the rule against killing someone whose probability of being guilty is

Constant writes:

Is the proposed rule that it is always right to shoot someone whose probability of guilt is > 50% but otherwise not?

Suppose that I possibly (possibly not) shoot someone whose probability of guilt in a previous murder is only 45%, but then at my trial the evidence against me only comes to 55%. So from the jury's point of view, there is a 55% probability that I killed the other person, but in that case there is still a 45% probability that I killed a guilty man. So from the jury's point of view, I probably did not kill an innocent man. That is, the probability that I shot an innocent man is 55% times 55%, or about 30%.

Am I to be found guilty and punished or not? It depends on what the rule is. There is a 55% probability that I am guilty of violating the rule against killing someone whose probability of being guilty is

Isn't the whole point of the preponderance rule to avoid such a situation? And yet it appears to produce the same situation.

Recall the reasoning employed in the defense of the preponderance rule. The father, it is noted, has violated (or is about to violate) the reasonable-doubt rule - that is, the rule against killing people whose probability of guilt is less than 90%. For this the law says he himself is a murderer and thus should be executed. This law is critiqued from a standpoint outside the law. We are asked to consider the point that if the father shoots the defendant, then the probability that he has killed an innocent man is only 20%. I believe that this is the gist of the argument, expressed in the words "the man who probably is not a murderer". The "probably not" refers to the 80% probability that the person he shoots in fact killed the child, and thus to the 20% probability that he has shot a person who is innocent.

So the argument is conducted by stepping outside of the rules and critiquing them from that external standpoint. The rule that is stepped outside-of is the rule against killing people whose probability of guilt is less than 90%.

If this is an allowable move, then it is allowable for me to make the same move against the preponderance rule, as I believe that I have done above. In the scenario I have sketched, there is only a 30% probability that I have shot a man who is innocent, but there is a 55% probability that I have violated the preponderance rule (which is a rule against killing people whose probability of guilt is less than 40%). Thus the preponderance rule produces the same kind of injustice as the "reasonable doubt" rule, which can be critiqued from the same external standpoint.

Constant writes:

I don't know what's happening. First it cut off my comment in the middle and then the correction duplicated the initial fragment. I'm going to leave things as they are because I think there's at least one complete copy. I must have done something but I don't know what.

Constant writes:

I thought about it. I know what happened. I used a less-than sign, which was interpreted as the opening of an html tag.

Richard Phillips writes:

The objection is that your logic creates incentive problems for future potential avengers. By the law of large numbers, we would be indulging to private psychological benefits of revenge at the expense of lives of some innocent people.

Richard Phillips writes:

The objection is that your logic creates incentive problems for future potential avengers. By the law of large numbers, we would be indulging to private psychological benefits of revenge at the expense of lives of some innocent people.

R Phillips writes:

The objection is that your logic creates incentive problems for future potential avengers. By the law of large numbers, we would be indulging to private psychological benefits of revenge at the expense of lives of some innocent people.

reason writes:

What sort of court allows emotionally involved participants in with guns?

And of course your argument is an argument for non-lethal but disabling technology instead of the primitive ballistic devices now in use.

Barbar writes:

We can construct hypothetical arguments against any legal system.

Let's take the facts in this case, but presume that the defendant is in fact innocent because the father is mistaken (eyewitness testimony is often wrong). Oh snap, the reasonable doubt standard works brilliantly! And hm, is it a tough call to stop a certain murderer to protect a non-murderer? Gee, not at all.

The heights of rationalist thinking. Yes, let's scale back democracy and put Bryan in charge.

tim writes:

"Only the preponderance rule lets us escape the bizarre conclusion that we should kill the man who is less likely to be a murderer in order to save the man who is more likely to be a murderer."

This is the sort of comment that gives economists a bad name. What if the jury had found the accused guilty, and the father still pulls a gun to shoot the man? Clearly it is not the preponderance rule that leads to the 'bizarre' conclusion that the bailiff should carry out his sworn duty and prevent the father from killing.

In this scenario, shouldn't the father shoot the members of the jury too? This would provide some 'costs' that might prevent future jurors from making the wrong decision.

But, now, let's suppose the jury found an innocent man guilty, and the defendant was wrongly sentenced to death. Shouldn't the bailiff shoot the members of the jury? After all, this would provide incentives for future jurors to be more careful and get their decisions correct.

Horatio writes:

Assuming the father did indeed witness the accused murder his daughter, he has the right to do anything he wants to him. However, the bailiff is also justified in shooting the father. Each is acting on the information they have.


If someone burns my car, I have the right to demand compensation for the value of the car as well as the trouble of getting a rental and shopping for a new one. Those are my property rights. If I wish, I can also forgive the criminal. But what if I died before the criminal could compensate me? My next of kin would inherit my property rights and would be justified in demanding compensation from the criminal. Now what if I was murdered? Again, my next of kin would inherit my property rights, which now include complete ownership over the individual who murdered me.

Dan Weber writes:
Why doesn't the legal system weight punishments according to confidence in the verdict? The jury is asked two questions: "Guilty?" and "What punishment?" They are asked to answer each question without regard to the other.

Juries generally do have that latitude. If they're pretty sure of a first-degree murder, but not past the "reasonable doubt" test, they can compromise and choose to convict for second-degree murder.

This requires, however, that the prosecutor has included second-degree murder charges among his list of charges.

If I were to try and contort this into the hypothetical of the original post, I would have the father shoot the prosecutor for not giving the jury a way out. Chew on that one! :D

KSH writes:

Are you sure you want to make it easier to convict?

Matt writes:

Why doesn't the bailiff shoot the gun out of the father's hand?

Every crime is a separate event. Consider this scenario: a murderer/child molester gets 20 years in prison, and a member of the Aryan Brotherhood convicted of assualt is killing him in the yard. The guard has a high powered rifle and there is a crowd, his only shot is a kill shot. If he had time to think, might think, "One of these guys is going to die. No one in this state gets the death penalty. Therefore, I can allow a child rapist/murderer to die, or I can kill a thug/murderer to be."

Nacho Bizness writes:

The father was an eyewitness to the original crime and he actually knows what he saw. Attempting to shoot the defendant in the courtroom may cost him his own life, and even his best case scenario will find him on trial for murder. It is highly unlikely that he would take this action unless he were, in fact, telling the truth in his testimony that the defendant did commit the crime. The bailiff knows this; the jury could not. Most likely the bailiff is now convinced that the defendant did in fact kill the girl, and that the father is correcting a grave miscarriage of justice, so he will only arrest the father after the fact and not try to stop him.

Dwight Schrute writes:

[Comment deleted for supplying false email address. Please email the webmaster@econlib.org to request restoring this comment.--Econlib Ed.]

Jason Briggeman writes:

A preponderance standard (vs. a reasonable-doubt standard) would have much higher returns to the manufacturing of evidence, i.e., framing. Consider a defense team able to create reasonable doubt, but a much better funded prosecution team able to generate mountains of testimony and evidence -- much of it tangential or circumstantial, perhaps, but any given jury of 12 random people may well interpret "preponderance" to be related to quantity and simultaneously fail to discount for quality or investment.

Bill writes:

I confess I don't see the problem. For someone who believes in the rule of law and individual liberty, it would seem to me that the suspected acquited murder would have his liberty much impaired by being shot. And if the government, as measured by the bailiff's actions, can't be trusted to protect one's liberty, then what good is the government?

I guess the problem here is that one person's liberty being upheld necessarily means another person's liberty is damaged. It really is a zero-sum game in regard to this transaction, and there are many such transactions in society today. Gun ownership, allocation of tax dollars, laws which are designed to protect existing property rights, higher taxes on higher earners, etc. Here again I would suggest the best approach is to consider the individual, the collective, and the relationship between both.

I would further suggest that if one follows libretarian beliefs in invidual liberty and freedom too far, without stopping to consider the collective, you end up with the feudal state, where property rights are absolute, but only as good as the force you have to protect them yourself.

Scott Scheule writes:
We can construct hypothetical arguments against any legal system.

Let's take the facts in this case, but presume that the defendant is in fact innocent because the father is mistaken (eyewitness testimony is often wrong). Oh snap, the reasonable doubt standard works brilliantly!

As would the preponderance of the evidence standard.

The heights of rationalist thinking.

Don't be an ass.

Comments for this entry have been closed
Return to top