Voter Fraud: Type I & Type II Errors.

Caveat: IANAL, I am ignorant of the law, however, I know a little about statistics and testing hypotheses.

On one hand:

Suppose we are confronted with making a decision based on limited data; most real world data is incomplete. Consider, for example, the charge that Ted murdered Jack

Our legal system doesn’t require that jurors be omniscient. It only requires that jurors begin by assuming that Ted is Innocent until Proven Guilty. In more precise language, jurors should begin with the hypothesis that Ted is not guilty as opposed to the alternative that Ted is guilty. Jack is guilty is not a reasonable alternative hypothesis. 

We do not test whether Ted is innocent versus Ted is not innocent. If the reason for this is not immediately obvious, I hope it soon will be. The jurors examine the presented evidence, and ideally without bias, evaluate the accumulation of evidence and arguments.

If the accumulated arguments and evidence presented by the prosecution persuades the jury beyond a reasonable doubt, then the jury should reject the hypothesis that Ted is not guilty, and accept the alternative hypothesis that Ted is guilty. On the other hand, if the defense persuades the jury that the evidence is insufficient and instills a reasonable doubt, then the jury should not override the starting premise, the hypothesis that Ted is not guilty. Note that a finding of “not guilty” does not declare that Ted is innocent, only that the evidence did not rise to the level of beyond a reasonable doubt in order to convict.

Rising past the level of a reasonable doubt is not a precise standard because it suggests that what is reasonable must be decided by individual jurors. This is one reason that trials end up with hung juries or have decisions second-guessed by critics. That is also why we have legal opinions in advance of trials rather than facts.

If you are on the jury, then you must make a decision based upon your interpretation of the “facts” as presented to you by the prosecution and the defense and your background and bias. Here are some possible decisions you could make:

Suppose Ted is not guilty,

1.       You vote not to convict. You made a good decision.

2.       You vote to convict. You made a bad decision, an error (TYPE I).

Suppose Ted is guilty,

3.       You vote to convict. You made a good decision.

4.       You vote not to convict. You made a bad decision, an error (TYPE II).

Statisticians describe decision 2 as a TYPE I Error: the rejection of a true statement. In this case, the hypothesis is that “Ted is not guilty.” The TYPE I Error here results in Ted, who is not guilty, going to jail. The purpose of the reasonable doubt standard is an attempt to minimize the frequency of the TYPE I error sending “innocent” people to jail.

Statisticians call decision 4 a TYPE II Error. In this case, a guilty person is released rather than incarcerated. Notice that if Ted is guilty, and the jury makes a TYPE II Error, then the actual judgement is not that Ted is innocent, but that the evidence was insufficient to convict. Hence Ted is declared “Not Guilty” rather than innocent. Still Ted goes free.

Folks get incensed when they believe that guilty folks get declared not guilty because of the beyond a reasonable doubt standard. They want to make it easier to convict. For example, in civil suits, a plaintiff wins by a preponderance of the evidence. The winner in court receives the entire judgment if just 51% of the evidence supports his side. Some folks don’t believe that is fair either because civil cases may exhibit that both sides share the blame, and giving someone 100% of the judgement because their share of the blame is less than 50% doesn’t seem fair.

Moreover, if the preponderance of the evidence were the standard for conviction in criminal cases, there would be a lot more convictions and hence a lot fewer TYPE II Errors. However, this would also result in a lot more “innocent” folks getting convicted, and thus, a big increase in Type I errors.

In fact, TYPE I Errors and TYPE II Errors are interconnected. Change the standard for making decisions to decrease TYPE I Errors, and TYPE II errors increase. Change the standard to decrease TYPE II Errors and TYPE I Errors increase.  In jury terms, make convictions easier and more innocent people go to jail. Make it more difficult to send innocent people to jail, and more guilty people go free.

On the other hand:

That brings me, at last, from murder to voting.

Despite absolutely no evidence of widespread or systematic voter fraud, some folks want to make it more difficult to vote. Here are the possible outcomes in “tightening” voter qualifications and rules to vote.

Suppose that Mary should NOT be eligible to vote:

1.       The new law keeps Mary from voting. A correct decision.

2.       The new law allows Mary access to voting. An incorrect decision. A TYPE I Error.

Suppose that Mary should be eligible to vote:

3.       The new law allows Mary access to voting. A correct decision.

4.       The new law keeps Mary from voting. An incorrect decision. A TYPE II Error.

I submit that many of the new laws being proposed in numerous states are solutions in need of a problem. No reliable data provides any evidence of systematic voter fraud. Sure there are scattered individual cases of attempted fraud or misuse of the system, but the amount is so small as to be statistically insignificant. Moreover, there is plenty of evidence of widespread campaign contribution fraud and income tax fraud, but that doesn’t seem to engender any reformation of the laws governing those activities.

Many of the voter law changes are not really about fraud but rather about making it more difficult to vote, including for many people who should be eligible to vote. The unstated purpose, except when some politician slips up and answers the questions truthfully, is to increase the likelihood that some folks who qualify to vote, will not be able to vote.

The stated claim is that the laws cuts down the likelihood of a TYPE I Error with a goal of having zero folks vote who are NOT eligible. But attempting to reduce a TYPE I Error by moving towards no fraud means increasing the the TYPE II Error and thus increasing the number of legitimate voters who will be kept from voting. However, this side-effect accomplishes the unstated purpose.

Why would anyone want to suppress voting when the frequency of voter fraud is astonishingly low?

1.      To appeal to conspiracy theorists and a misinformed base (It doesn’t matter if there was no fraud, our base has no faith in the system, hence we must do something.)

2.     To prevent people who don’t vote the way we want them to vote from voting. (Some folks are not thoughtful voters. In other words, they don’t vote the way we want them to vote.)

On the third hand:

Consider a case in point.

According to the Guardian, “Four years ago, (Crystal Mason) was on supervised release, similar to probation, for a federal felony conviction related to tax fraud. She didn’t know that Texas prohibits felons from voting until they finish their sentence entirely. Mason voted in the last presidential election at the urging of her mother and cast a provisional ballot when poll workers couldn’t find her name on the voter registration rolls. The ballot was never counted because Mason was not an eligible voter.”

The ballot was never counted because the voting system worked the way it is supposed to. Ms Mason didn’t know she was not permitted to vote. She did not lie about who she was. As required for provisional ballots, her status was investigated, and her ballot was declared ineligible and not counted. She did not knowingly commit fraud. Nonetheless she was arrested, convicted, and sentenced to five years in prison, and had her eleven months early release revoked. On appeal, a judge ruled that it didn’t matter that she didn’t know she wasn’t allowed to vote, or that the folks at the polls offered her a provisional ballot that would only be counted if she were later found eligible: she should have known.

Voter intimidation. Even if someone is pretty sure they are eligible, why would they risk a provisional ballot after this story made the news?

All this took place under older voting laws in Texas. Apparently, voter intimidation is not enough. Currently, the Texas legislature is ramping up “improvements” to keep up with the avalanche of proposals in other red states.

Suppose you were in a park and wanted to discard a coffee cup. Two cans are there for disposal. One says Garbage, the other says Recycle. You ask a distracted park ranger which to use, and he says, “I’m on the phone. Set it on the bench, and I’ll put it in the correct can when I finish.” You set it on the bench next to the cans. As you turn to leave, a whistle blows, and another park ranger rushes in and arrests you for littering.

In 2013, former Speaker of the House, Jim Wright was denied a photo ID for the purpose of voting by the State of Texas. After bad national publicity, the state relented. If someone denies you the right to vote when you are in fact entitled to that privilege, you can never be reimbursed for missing the opportunity to vote.

The people pushing for tighter voter registration laws to deliberately disenfranchise qualified voters, a TYPE II Error, by appealing to conspiracies, fear, and lies are not your friends. They do not have your best interests at heart.