A Typical "Voter Fraud" Example

Today, the 28th of August, 2013, is the 50th anniversary of Dr. Martin Luther King Jr.'s famous "I have a dream" speech, which contains this line:

We cannot be satisfied and we will not be satisfied as long as a Negro in Mississippi cannot vote and a Negro in New York believes he has nothing for which to vote.

The movement this speech came from and drove onward eventually led to the 1964 Civil Rights Act and the 1965 Voting Rights Act, both of which in large part dealt with ensuring the right of all American citizens—especially minorities, and African Americans in particular—to vote.

Everyone knows this, and today, everyone is celebrating this. And yet, on this very day, the states of Pennsylvania, Texas, and North Carolina are in court trying to defend the removal of these same rights, from these same citizens. The states claim, of course, that undoing half a century of progress isn't the objective of these laws. The laws being challenged in court are instituting requirements for photo identification at the time of voting, and the justification for them is to prevent fraud. So today I'm going to walk you through an example of exactly how accurate these claims of fraud are, from a state that isn't—yet—in court: South Carolina.

We're using South Carolina because advocates for this sort of law made a specific claim, made it very publicly, and it was investigated very extensively. We begin with Department of Motor Vehicles Director Kevin Schwedo, and Attorney General Alan Wilson, who claimed that they had a list of 953 deceased individuals whose names had been used to fraudulently vote in the 2010 election. The South Carolina Law Enforcement Division (SLED) spent nearly 7 months examining the list, and this is what they found:

• -746: Hadn't voted in the 2010 election. Only by including 74 elections over a 7 year period could that number be reached. SLED wasn't going to go digging that far back, but was happy to investigate the remaining 207.
• -92: Junior/Senior name recognition errors. In other words, John Smith Sr. had died, and John Smith Jr. had voted, legally.
• -6: "Clerical errors," where the poll worker had marked the wrong person in the poll book and not fully erased their mistake.
• -56: "Bad data matches," where the state had the wrong social security record on file for the voter. The owner of the SSN had died, but the vote was legit.
• -5: Cases where election workers incorrectly marked the wrong person as having voted absentee.
• -3: Absentee ballots which the state had issued in the wrong name, but the (legal) voter who requested it filled it out anyway.
• -32: Scanner errors. The machine screwed up.
• -3: Voters who were alive when they requested their absentee ballot, but died before the day of the election.
• That left -10 cases that needed further scrutiny:
  1. Jr. voted, father deceased. (Although Jr. hadn't registered to vote.)
  2. Agent exhausted all leads, no further information.
  3. Jr. voted, legally; father deceased.
  4. Jr. voted, legally; father deceased.
  5. Agent exhausted all leads, no further information.
  6. Agent exhausted all leads, no further information.
  7. Jr. voted, legally; father deceased.
  8. Unrelated individual with the same name, who was a registered voter, voted. (Although in the wrong precinct.)
  9. Jr. voted, legally; father deceased.
  10. Jr. voted, legally; father deceased.

All told, that's at most 4 fraudulent votes, 5 if you count the guy who went to the wrong precinct, but saw his name in the book and voted anyway. And statistically, the 3 inconclusive cases were probably not fraud either. The only certainly fraudulent vote was the guy who never registered and voted under his dead father's name. (Which I guess they let slide, because SLED didn't file charges against anyone; I'm sure Jr. will either register or stop voting now that he's been investigated and let off with a warning.) That's 1 out of 207, or less than half of a percent, or 1 out of 953 if you use the initially very-well-advertised and very-inflated number, which is hardly more than one tenth of a percent. And this in an election where 1,365,480 votes were cast. Meaning this "huge" example of voter fraud resulted in, literally, less than one fraudulent vote in one million. And this is how it always goes with these cases.

Further more, I would like to point out that many of these claimed cases of fraud were by absentee ballot, which voter ID laws would not be able to prevent. A great many more were cases where the voter would have had photo ID with the correct name, including the one and only certainly fraudulent one, so a voter ID would not have stopped these either. Almost all the rest were errors on the part of the state, which voter ID also can't stop.

But maybe you're thinking "Okay, none of these cases were fraudulent, but there's surely some other cases that were. And even one vote, one fraudulent vote, is worth stopping; we're talking about our democracy after all!" And now we're into cost-benefit analysis. Because voter fraud does—rarely—happen. But if stopping one fraudulent vote is good, not incorrectly preventing thousands of legitimate votes must be even better. The ACLU of Pennsylvania was able to, almost instantly, find 5 voters just in the Philadelphia area who would be unable to meet the state's voter ID requirement, even though they are clearly legitimate voters, and South Carolina has 3 times as many people as Philly. Imagine how many such voters must be in South Carolina, if you look even a little bit harder. If it's just two, then the voter ID law is a net-negative. And it is certainly more than two, and possibly as high as 100,000.

I know that this issue has become a partisan one (although some Republicans have defected from the party line) and it pains me to bring a partisan issue to a blog that I want to be welcoming to all advocates of democracy; left, right, center, and off-axis. But I can't look at the arguments being made by the other side and see anything other than barely-concealed discrimination, because those 100,000 voters who will have difficulty getting an acceptable ID are not evenly distributed across the population, they disproportionately come from minorities, and African Americans in particular. And I will not let that go unmentioned on this golden anniversary day. We should, at all times and in all ways, do everything we can to get as many people as we can to vote. That forms the bedrock of our electoral system; without that, the calls I make for better voting methods are pointless.

The Tweet of all Evils, or The Least of all Tweeters?

I'm a huge RSS fan, but with the death of Google Reader last month (I'm using Feedly now, by the way) it's well past time that I actually put this blog through to Twitter (which I guess is how the cool kids do everything these days?) If you've been with us long enough (long enough that you recognize where the user icon came from) you'll remember that we tried this once before. But this time I've set things up (through TwitterFeed) so that new post will automatically be tweeted (one has already gone through, so I guess it's all set up correctly; yay!)

So if that sounds like fun, you can easily catch all the new posts to The Least of All Evils by following @LeastOfAllEvils. I don't expect to be much of an active Twitter conversant through that account though, so if you want the full experience (which is to say, a bunch of stuff about politics, but also stuff about programming, random science, board- and video-games, and life in Anchorage) you can catch me at @Mudlock.

Egypt Didn't Want This

It sounds ridiculous to have to say it. Of course Egypt didn't want this. No one wants death and violence in the streets. But it does seem that a majority of Egyptians support the military coup that recently removed their first democratically-elected president, when hardly one year ago a majority of Egyptians supported the election of that same president. Or did they? I'm not suggesting nefariousness, I'm not suggesting ballot-stuffing, election fraud, vote buying; nothing like that. No, Morsi won the election completely within the rules. But despite that fact, a majority of Egyptians did not, in fact, support him. The problem was, the rules were incredibly stupid. Un-democratically stupid. Self-defeatingly stupid. And we could shrug and say "I guess that was dumb of Egypt," except that we use the same rules in much of the United States.

The rules for Egypt's 2012 presidential election (and for most seats of the preceding parliamentary election) were simple: plurality, and if no candidate receives a majority, the top-two candidates head to a runoff. Now, there were five candidates who received more than 2% of the vote in the first round of the election. And it is entirely possible—and I would say it's probably—that the first- and second-place candidates were the two least liked candidates of that group, by which I mean, either one would have lost in a runoff election against any of the other three. In other words, in 2012, Egyptians elected their second-least-preferred choice to be their president. In such a revolutionary time, with years of political oppression finally released, and with several vastly different ideas of where to take the country politically, is it even a surprise that such a candidate would be removed less than a year later?

I've constructed a simple example using Yee's Voteline simulator that roughly parallels the results in Egypt. Beginning with Morsi (a Muslim Brotherhood leader and the election winner) on the left, then Abou al-Fotouh (a former leader in the Muslim Brotherhood), followed by Sabahi (a secularist and opponent of the Mubarak regime), Moussa (one time Minister of Foreign Affairs under Mubarak, but also something of a rival of his), and finally Shafik (named Prime Minister by Mubarak on the eve of his expulsion.)

When the middle three candidates are removed, the result mirrors the outcome of the runoff election. It's a very approximate recreation, and I'm not entirely comfortable with Sabahi and Moussa's positions in particular, but there's a rough line from most-recent leader of the Muslim Brotherhood to most-recent member of the Mubarak government, with Sabahi (having not ever been a member of either) in the middle. Now, I'm no expert on Egyptian politics, and a one-dimensional simulation like this must necessarily leave out some details, but I believe my result to be at least illustrative if not entirely accurate. You're more than welcome to play around with Yee's simulator yourself to see if you can find something better.

I found it interesting, while I was reading up on the candidates, that most Egyptians voted for a secularist candidate (Sabahi, Moussa, or Shafik) and also most Egyptians voted for a candidate unaffiliated with Mubarak's government (Sabahi, Abou al-Fotouh, or Morsi) but the runoff election pitted the most-religious candidate against the candidate closest to Mubarak. In other words, the 3rd-5th place candidates split the middle of the electorate, dooming the runoff to extremism. You can also see that most other election methods, including approval voting, Borda count, and any Condorcet method, would all have elected Sabahi (however, this is where I lose some faith in the model, as I think Moussa may have proved to be the true winner under some of those methods, but it would take a multi-dimensional model to see it. A Sabahi/Moussa runoff would have been a good election.) Interestingly, instant runoff voting ends in the same Morsi/Shafik showdown as plurality plus runoff, and ends the same way, further condemning that method.

The moral of the story is: Winning an election, if that election has stupid rules, doesn't necessarily get you the best result. In a developed democracy, we can roll with it, wait four years, and try again. In a high-stakes post-revolutionary environment, getting it wrong this way leads to violence, death, and an irrevocable (if unfair) stain on democracy itself. But democracy didn't fail in Egypt. Democracy didn't kill 600 Egyptians last week. Plurality voting did.

Yee Paper on Visualizing Voting Methods

Just a quickie (I know, I haven't posted that Egypt piece I promised; turns out, just because you don't know a lot about a foreign country's politics, doesn't mean they're simple) and not something that's new if you've been following us closely, but Ka-Ping Yee has a nice, simple, short paper based on his previous work, titled "Disadvantages of Transferable Vote Systems". This one is restricted to just plurality, IRV, and approval, and comes out strongly for approval. Also, the pictures look pretty sweet, so check it out.

Lessons From Egypt: The Voting System Matters

I'm writing a longer piece on this same topic, but I want to get something out while it's still in the news.

• First, David Rea writes "Egypt has not rejected Democracy," which lays out the problem and a possible solution. Short version: when an organized but disfavored minority wins an election, that's bad; for the nation, for the voters, and for the reputation of democracy itself. Use approval voting.
• Second, Warren Smith provides the numbers to back those claims up. A better voting system (such as score voting) probably would have chosen Amr Moussa, who came in 5th(!) in the actual election.

Elections matter. Getting them right, matters. Getting them wrong can lead to violence, war, and death.

LoAE Mascot: The Least of All Weasels

From frozenplanet.tumblr.com

Too much serious business last month, so it's time for some relaxation. I've been lying in wait with this post for a few weeks now, waiting for the perfect time to pounce. You see, all the world's great organizations—sports teams, political parties, even nations—have something in common, and that's mascots. So after careful consideration of several possibili... oh, who am I kidding? It's a silly pun, plus look at how cute this little guy is!

This is Mustela nivalis, or the least weasel. Found in North America from the U.S. eastern seaboard to Alaska (much like myself) as well as across northern Asia, all of Europe, and even into northwestern Africa, the least weasel is astoundingly prolific, but rarely noted, possibly because of it's small stature (weighing just 1 to 9 ounces (30-250 grams.)) Which is a mistake, since this tiny carnivore can take down prey five times its own size, and more! Legends even say that the weasel is the only creature capable of defeating the basilisk. Basically, they're fuzzy little magical four-footed ninjas, and who doesn't love ninjas?

I know who has my vote for next King of the Animals.

From Bering Land Bridge National Preserve

Red Crayon Statistics

I know, I implied the previous post would be the last one about Minneapolis. But I can't help it. My first post on the subject implied that former interim elections official Patrick O'Connor was lying about the 2009 ranked-choice-voting-powered election in Minneapolis, as the official report (which he commissioned) contradicts his claims. But today, the exact same lie is being propagated by the author of the report!

I started to question my own sanity here. These guys really should know better, and no one could really expect to get away with such an obvious lie. Is it possible I'm the one who's wrong? But then I remembered how absolutely ludicrous it is to suggest that only one ballot out of nearly 50,000 could be spoiled. No system has ever been that accurate, and anything under 1% is considered a triumph of electoral vigilance.

But then I noticed the subtlety they're using here. Both O'Connor and Schulze talk about spoiled ballots earlier in their articles, but when they try to counter the complaints of excessive ballot spoilage under RCV, their specific claim is that only one ballot was "uncounted". Now, given the context of that claim, you might assume that "uncounted" and "spoiled" are being used as synonyms for each other. Nope! Let's check the report again!

Of the 45,968 total ballots cast, there were 1,888 spoiled ballots and 2,958 voter error ballots (2009 Election Statistics, n.d.), which indicates ballots with voter errors specific to Ranked Choice Voting, including overvote, repeat candidate, skipped ranking and undervote (Minneapolis Method for Hand-Counting RCV Ballot Sorter & Counter, n.d.). Comparatively, during the general municipal election in 2005, there were 755 spoiled ballots of the total 70,987 absentee and in-person voters (Voter Turnout and Registration, 2005). In 2009, there was only one ballot cast that was totally defective and not counted. This was a ballot where no ovals were filled in and a handwritten essay written in red crayon on it. Excluded from this analysis are any ballots not counted due to failure to comply with the rules regarding absentee ballots. These ballots were not opened to ascertain voter intent.

That is the "one ballot" they're talking about. There were 1,888 spoiled ballots, a more than 380% increase in spoilage rate over the previous election, but never mind that! Because only one guy wrote on their ballot in crayon that year, that means those 1,888 don't really count against anything.

The truth is, RCV nearly quadrupled the number of spoiled ballots in Minneapolis, and there would have been a ten-fold increase had the election been machine-counted rather than hand-counted. That alone aught to be enough to disqualify RCV as an election method, but there are plenty of other reasons for it, too.

I'm upgrading O'Connor's and Schulze's "damned lie" to "statistic".