Would Gessler Have Won the GOP Primary with Approval Voting?

gesslerOn Tuesday Colorado Republicans selected Bob Beauprez to run for governor—again. Queue “both ways Bob,” queue the “war on women.” (I doubt the Democrats will make much of Beauprez’s 2007 support for an insurance mandate, or his 2000 support for anti-gun laws.) I predict that he will lose—again (this time to John Hickenlooper, the incumbent). (This is no courageous prediction; I don’t think any of the candidates would have been able to beat Hickenlooper, despite his mishandling of the gun issue. I could be wrong, of course; general antipathy toward Democrats this time around could swing the governor’s race.)

What’s interesting about the primary vote is that four strong candidates split the vote relatively evenly. With 98 percent of the votes reported, the Denver Post offers the following results:

  • Bob Beauprez: 30.3%
  • Tom Tancredo: 26.6%
  • Scott Gessler: 23.2%
  • Mike Kopp: 19.8%

In other words, fewer than a third of Colorado Republican primary voters, or a little over 111,000 people (in a state with a population of around 5.3 million people), cast a vote for Beauprez—hardly a popular uprising.

Consider what might have happened under approval voting. The basics of approval voting are straightforward: Each voter gets to vote for as many candidates as he or she “approves” of. The candidate with the most votes wins. For example, if I had voted in this primary under approval voting, I would have cast a vote for both Gessler and Tancredo (despite my deep disagreements with the latter).

Although it’s quite possible that Beauprez would have won under approval voting as well, I think there’s a good chance Gessler would have won.

Here’s my reasoning. Beauprez is the milquetoast, establishment candidate, and I think a lot of people voted for him just because he’s tall and grandfatherly, he has congressional experience, and he’s not as ornery as Gessler (a quality of Gessler’s I find appealing) or as weighed down by baggage as Tancredo. (Kopp was never a leading candidate, despite his incessant YouTube ads.) I think that, under approval voting, many people who voted for Beauprez also would have voted for Gessler. I think that many people who voted for Tancredo also would have voted for Kopp, and vice versa, but that many people who voted for those candidates also would have voted for Gessler as their second choice.  And I think that a disproportionate number of people who voted for Gessler would have voted for a single candidate. In this scenario, Gessler may well have pulled ahead.

Of course, there’s no way to know for sure. The only thing we can know for sure at this point is that fewer than a third of Republican voters cast a vote for the Republican nominee for governor, and that doesn’t give me much confidence that the outcome accurately reflects voters’ preferences.

Those who have a different guess as to what the outcome would have been under approval voting are welcome to explain their reasoning in the comments.

Incidentally, Mike Dunafon is also running for governor, as an independent (and I may well vote for him), and there’s also a Libertarian in the race. Although Hick might get more than half of the total votes, he may well win with Beauprez and Dunafon (and the Libertarian) combined earning more than half the votes. If that happens, I will take the opportunity to write yet another post about the benefits of approval voting.

Westminster Should Consider Approval Voting

I read in the January 25 Westsider that the Westminster City Council is considering eliminating run-off elections for mayor:

If the top candidate does not receive 40 percent [under the current system], the top two candidates face off against each other during a run-off election. This process requires a second election, costing about $100,000. . . . Resident Tim Kauffman told council [at a January 14 meeting] the run-off election is important because the mayor position needs widespread community support.

The council will decide the measure to eliminate the run-off election on January 28, the paper reports.

But the city could avoid both problems—a costly run-off and a low-popularity mayor—simply by instituting approval voting, a process I wrote about a couple years ago.

Here’s how it would work. For the mayor’s election, voters would see all the candidates’ names on the ballot. Voters could vote for one or more of these candidates—as many as they “approved” of. Then the candidate with the highest vote total wins.

This guarantees that the winner has broad support, yet it saves the cost and hassle of a run-off election. What’s not to like?

New Group Seeks Voter Integrity

At Monday’s Liberty On the Rocks Flatirons event, Jeff Kelly discussed his new group, “Colorado Voter Protection.”

Kelly said his group’s goals are three-fold: first, to clean up the voter rolls; second, to “recruit and deploy honest, trained poll watchers”; and, third, to prevent voter fraud.

Regarding that last point, Kelly said he’d like to see legislation requiring the verification of citizenship, minimizing mailed ballots, and requiring voter identification.

Gessler Addresses Anonymous, Verifiable Voting

Secretary of State Scott Gessler discussed the standards of anonymous, verifiable voting when I interviewed him at the Independence Institute’s annual banquet. He said that, while current practices are “solid,” potential improvements in technology might further improve the system.

For additional videos from the II’s banquet, see also:

Kopel: ObamaCare Mandates Unconstitutional

Independence Institute Banquet

Why Voting Integrity Requires Proof Positive

Recently I wrote, “Mailing out ballots to inactive voters is an open invitation to voter fraud.”

This prompted leftist gadfly Alan Franklin to reply in a series of Twitter posts, “[I]f mailing ballots to inactives is ‘an open invitation to voter fraud,’ I assume you have… evidence of that? … Because if you have no evidence… you’re just another clown using baseless scare tactics. You know that, right?” He further claimed that I’m using “baseless scare tactics to stop ballots from going to voters” and “[t]echnicalities to suppress voting.”

This is a serious topic that deserves serious consideration (as opposed to Franklin’s ad hominem attacks).

The principle involved here is that the onus of proof lies on the one making an assertion. So if someone were to falsely accuse Franklin of theft, he could sensibly reply, “You have no evidence that I’ve committed theft; therefore, your accusation is groundless.”

Does a voting system likewise require positive proof of fraud before that system may be declared unsound and prone to fraud? No. The difference is that voting is the positive evidence required for an election result, and that positive evidence must be collected and presented according to reasonable standards of integrity.

To return to the example of falsely accusing someone of theft, imagine the following discussion:

Ben: “You are a thief”
Alan “No, I’m not. You have no evidence of that, and you are lying.”
Ben: “But you have no evidence that I’m lying, so therefore you are a thief!”

The problem is that Ben bears the burden of proving that Alan is a thief. Ben cannot throw the burden of proof back onto Alan.

Likewise, the voting system bears of burden of proving which candidate (or side) earned the specified number of votes. One cannot reasonably accept just any system of voting based on the claim, “Well, you have no evidence there’s anything wrong with this system.” Voting requires positive evidence that it yields accurate results.

Imagine some different scenarios that illustrate this principle.

In Country P, the dictator says, “I am the rightful ruler of this country, because the people support me!” If a critic were to answer, “There’s no good reason to think the people support you; why not throw the question up for a fair and open election?” What we we say if the dictator answered, “I know my people, and I know they support me. You have no evidence that there’s anything fraudulent about my claim, you clown!”

But let’s say we could persuade the dictator that an actual vote by the reasonably qualified electorate would better determine which leader has the majority’s support. The dictator says, “I have collected all the ballots, and I have counted them myself, and I am the clear winner.” If the critic were to reply, “It’s not good enough for you to count the ballots yourself, because you have a strong incentive to miscount or ‘lose’ the ballots you don’t like. Therefore, the ballots must be counted in an open and verifiable way.” Again we would not be persuaded if the dictator replied, “You have no evidence that my vote count is flawed, you clown!”

Let’s switch the scenario to something more like our actual system. Suppose the Secretary of State issued the following proclamation: “I am standing up to the opposition’s voter suppression machine! I want to ensure that every qualified voter has an opportunity to cast a ballot. Therefore, I am mailing ten copies of the ballot to every residential address in Colorado, to ensure there are plenty of ballots to go around to all qualified voters. Moreover, people can also walk into my office, without any identification (because demanding ID would make me part of the voter suppression machine!), and request all the extra ballots they may need for themselves and others they know who are qualified voters.”

A critic might argue, “But, Secretary, such a system would be an open invitation to fraud!” What would we think of a Secretary who answered, “You have no evidence that such a system would result in fraud, you clown!” It would quickly become clear which party more closely resembles a clown.

A voting system requires positive proof that only qualified voters cast ballots, and that each voter casts only one ballot per election. You can’t just mail out ballots to everybody in the state. You can’t just let everybody wander in off the streets to vote, without identification.

Mailing ballots to inactive voters clearly fails the test of voting integrity. Often voters become inactive because they move out of state. Mailing ballots to those old addresses, absent any additional evidence that the voter actually lives there, is just asking for fraud. Setting up a system of voting that easily allows fraud is stupid and contrary to the very purpose of voting.

(Remember that inactive voters easily can reactivate themselves.)

Indeed, I worry that the entire strategy of mailing out ballots invites fraud. How many elderly voters get a little “help” from their activist grandchildren? How many people cast ballots that aren’t actually theirs? We do not know. It is impossible to know. And the absence of proof of voting integrity is the problem.

If we are going to maintain the system of mailed ballots, at least that system must include basic safeguards to prevent the worst and most obvious invitations to fraud.

Far better is to return to the polling place. (Note that there could even be a roving polling facility for the physically incapacitated.) Then voters can show their ID at the door. It’s clear that one voter casts one vote by his or her choice. That provides positive proof that the election is fair. Would that result in fewer people voting? Probably. But if you can’t be bothered to put in the minimal effort required to ensure voting integrity, seriously do you have any businesses helping to determine the future of our civilization?

Left Targets Gessler for Protecting Voting Integrity

Mailing out ballots to inactive voters is an open invitation to voter fraud. There’s no telling who’s going to receive and submit the ballot.

Therefore, Colorado Secretary of State Scott Gessler sensibly told Denver and Pueblo Counties that they should follow Colorado law and not mail ballots to inactive voters.

In response to Gessler’s protection of voting integrity, the left has waged a full-scale smear campaign against him, accusing him of racism and disenfranchisement. This is a major leftist strategy for the 2012 elections: smear all Republicans, conservatives, Tea Partiers, and free-market activists as racists, in order to raise sympathy support for Obama. It is a nasty, mean-spirited, intellectually dishonest tactic. If you want to understand why the left does this, read the article I coauthored on Saul Alinsky and Obama.

One basic issue is how an “inactive” voter can become “reactivated.” It turns out it’s trivially easy.

I called Andrew Cole about this; he is a spokesman for Gessler’s office.(Michael Sandoval has also written a bit about this issue.)

Cole said, “To become inactive, you have to have missed a general election, not responded to at least one follow-up postcard from your county clerk, and not voted in any subsequent elections such as a municipal election.”

So how does an inactive voter reactivate? Cole explained, “You can physically visit your clerk’s office, you can go to GoVoteColorado.com and with a state ID card activate your status online, and you could also write a letter to your clerk, but you’d have to have a signature to verify who you were.”

In addition, a previously inactive voter can get a ballot from the clerk directly “up until and including election day.”

So this is pretty simple: if you have not voted in a recent election, and if you have not bothered to reply to your clerk’s postcard, you can easily reactivate yourself online or by visiting or writing your clerk. Surely that’s not too much to ask of those determining the future of the free world.

While he was on the phone, I also asked Cole about the controversy surrounding military voters.

Cole said that issue arose in Pueblo County under clerk Bo Ortiz. Cole said Ortiz asked the Secretary of State’s office about the matter only this week.

According to Cole, Gessler’s office told Ortiz that he “can’t mail ballots to inactive voters.” However, he “should have resolved this issue weeks ago.” Cole said there are 80 inactive military voters listed in Pueblo County, and Ortiz has email contacts for 64 of them. In Cole’s words, Gessler’s office told Ortiz’s office this week, “You should immediately email all 64” with known emails and send postcards to the rest. Moreover, Ortiz “should have done that weeks ago.”

Cole summarized, “We’re asking Peublo County to follow the law, just like we’re asking Denver to follow the law.”

Apparently following the law is not a concept the left easily wraps its collective head around. Instead, the left would rather play the politics of smear and character assassination.

Denver Mayor Race Illustrates Benefits of Approval Voting

As of this writing, the top four candidates in the Denver race for mayor show the following vote totals:

Chris Romer: 28.5%
Michael Hancock: 27.1%
James Mejia: 25.7%
Doug Linkhart: 9.4%

The Denver Post reports, “If no candidate receives more than 50 percent of the vote in the May 3 election, the top two vote-getters advance to a June 7 runoff.”

But if you think about it, that’s a pretty foolish way to run an election. For example, what if Linkhart’s voters prefer Mejia over the other two? Too bad: Mejia is out (assuming the percentages hold). Or what if the voters of either Romer or Hancock far preferred Mejia over the alternative? Again, the outcome will be that a less-favored candidate will win anyway.

In addition, holding a runoff vote costs taxpayers more money (and taxes their patience as well).

The way to solve both problems — to pick the most-favored candidate and to do it with a single vote — is to institute approval voting. The basic idea is that voters can cast more than one vote. For instance, you could vote for Romer only, Linkhart and Romer, Romer and Hancock, or whatever other combination of candidates you think you could live with.

As it stands, chances are pretty good that the next Denver mayor will not be the candidate with the most support among the voters, though we’ll never really know. Even if the results turn out “right” in this case, inevitably the less-favored candidate will win in certain other races. I can think of lots of good reasons to institute approval voting, and no good reason not to.


Anonymous commented May 4, 2011 at 9:20 AM
Great post.


Anonymous commented May 6, 2011 at 6:04 PM
Why not prioritize by a 1 – 2 – 3 – 4? And maybe a NOTA?

Ari commented May 6, 2011 at 6:16 PM
One important reason not to let voters rank candidates is that it would be hard to process. If the numbers were hand-written, that would inevitably lead to disputes about the intended meaning of the handwriting. Plus, what if a voter did something like assign two #1s? A digital system also would be complicated. I suppose you could do it in waves: “Who is your first choice candidate, if any?” Then, of those remaining, “Who is your second choice candidate, if any?” And so on.

There’s another potentially important problem with rank (or instant-runoff) voting: a candidate who is broadly liked, who is everybody’s first or second choice, can get eliminated in the first round. I explain this here: http://blog.ariarmstrong.com/2011/01/atwood-pitches-approval-voting.html

Atwood Pitches Approval Voting

Frank Atwood promoted “approval voting” at a recent Liberty On the Rocks event:

While I was skeptical of approval voting at first, Atwood convinced me that it’s a good idea — even better than the “instant runoff voting” I’ve previously praised.

What is approval voting? You the voter can vote for any number of candidates on the ballot for a given position. For example, in the last gubernatorial race, you could have voted for both Dan Maes and Tom Tancredo, rather than just one of those candidates.

Now, in fact, John Hickenlooper won more votes than Maes and Tancredo combined, so it’s hard to claim that either Maes or Tancredo “spoiled” the race for the other candidate. (Democrat Hickenlooper won 50.7 percent of the vote, lackluster Republican Maes won only 11.1 percent of the vote, and third-party Tancredo won 36.7 percent of the vote.) But let’s consider a realistic possibility of a third-party conservative “spoiling” the race for the Republican or a Green candidate “spoiling” the race for the Democrat, meaning that if the minor-party candidate were not in the race, enough votes would go to the major-party candidate to make the difference.

In such cases, approval voting would allow a voter to approve both a major and a minor party candidate. Let’s consider a hypothetical three-way race with 100 voters under the two scenarios:

Chickenpooper gets 48 votes.
Tancledo gets 38 votes.
Maze gets 14 votes.

Under the “winner take all” voting we currently use, Chickenpooper (who we’ll assume is the left-leaning candidate) is the winner.

Chickenpooper gets 48 votes.
Tancledo gets 49 votes.
Maze gets 25 votes.

In this scenario some voters are voting for both Tancledo and Maze (so the total number of votes cast exceeds 100), and Tancledo emerges victorious. In this case, whereas Maze otherwise would have “spoiled” the election for Tancledo, under approval voting people can vote for Maze and still allow Tancledo to win.

Of course, in the real recent election, many people thought Maes was a complete joke, so some might have preferred both Hickenlooper and Tancredo over Maes and voted accordingly.

I favor approval voting because it allows minor-party participation without creating the risk of “spoiled” elections.

Why do I now think approval voting is better than instant runoff voting? Approval voting is both easier to implement and less prone to quirks.

Under instant runoff voting, a voter may rank candidates. For example, a voter could have picked Maes first and Tancredo second. Then, assuming Maes again came in third, all the votes that ranked Maes first and Tancredo second would go to Tancredo.

But think about ranking candidates in the voting booth. On a paper ballot, one would either have to write in numbers for the rankings or mark a candidate for the same race in successive votes. Electronic voting would require successive votes. But this process would confuse a lot of voters. For example, a voter may not realize that a second-place vote is not required.

Approval voting is a lot easier to set up. A voter simply marks a bubble (or the equivalent) for as many candidates as desired. It’s easy to understand and easy to implement.

It is true that any system of voting is subject to possible quirks; today’s quite pronounced quirk is that a minor-party candidate can “spoil” a race. Either instant runoff voting or approval voting would be less quirky. But I think approval voting would be best of all.

The Wikipedia entry on instant runoff voting suggests a problem which I’ll illustrate with the following scenario:

Suppose Lefty Lou is a fire-breathing leftist who excites his base but genuinely frightens much of the citizenry; for 34 percent of voters Lou is the first choice candidate. Suppose Righty Rick is a social conservative and the first choice of 35 percent of voters. And suppose Centrist Cal, a war hero and former NFL quarterback, is liked by pretty much everybody but the first choice of only 31 percent. Cal is the clear second choice for everybody who more strongly prefers one of the other two candidates. Let’s say that, due to imprecise polling, every poll shows the candidates within the margin of error, so voters have little idea who will actually win.

If voters simply voted their first choice under a winner take all system, Rick would win. But of course people can vote strategically under a winner take all system. Cal can argue, “Look, I know many of you are tempted to vote for Lou or Rick, but if you do that, you’ll risk putting your least-favored candidate in office. So vote for me!” And such an appeal may very well work.

But consider what happens under instant runoff voting. Assuming people vote their preferences, Cal is eliminated in the first round, throwing the election to Rick, despite the fact that, for 100 percent of voters, Cal is either the first or second choice. That seems like a bad outcome.

Under approval voting, Cal would easily emerge as the victor.

Nor am I persuaded by the critics of approval voting. For example, FairVote offers the following scenario:

To illustrate how approval voting violates majority rule, consider a primary with 100 voters and two candidates liked by all voters. 99 voters choose to approve of both candidates even though slightly preferring the first candidate to the second. The 100th voter is a tactical voter and chooses to support only the second candidate. As a result, the second candidate wins by one vote, even though 99% of voters prefer the first candidate.

The first problem with this example is that it is totally unrealistic. In a race with only two candidates, most voters would pick between the two based on their preferences. If someone were truly ambivalent between two candidates, voting for neither would have the same effect as voting for both. If 99 voters truly preferred the first candidate, many or most of them would vote that way. Does this involve a certain amount of strategy? Yes. But every system of voting does. The difference is that approval voting is more likely to generate a winner most voters can live with.

Another significant problem with the example is that obviously the 100th voter prefers the second candidate. Let’s say 99 voters barely prefer the first candidate to the second, and therefore vote for both candidates, not really caring which one wins. But one voter truly loathes the first candidate and therefore votes only for the second. Why should the very strong preferences of the one voter not outweigh the nearly-nonexistent preferences of the 99? But again this scenario is so implausible that we can safely ignore it.

Of course, if every race had only two candidates, a winner take all system would be fine (and a voter could approve of neither candidate simply by not voting in that race). The entire point of approval voting is to handle races with more than two candidates and thereby prevent scenarios of “spoiled” races.

Frank Atwood is right: Colorado should adopt approval voting.

Time for Real Choices at the Ballot Box

The following article by Linn and Ari Armstrong originally was published November 12 by Grand Junction Free Press.

If you think our Senate race was exciting — and we were surprised that Ken Buck lost despite his series of gaffes and oddball positions — just consider the drama in Alaska.

Lisa Murkowski held the Senate seat as a Republican, yet Joe Miller beat her in the primary. So Murkowski launched a write-in campaign, and the election remained under review as we submitted this column.

A November 3 AP story by Becky Bohrer explains the scope of the problem. Murkowski “was among 160 qualified candidates,” and “write-ins held 41 percent of the vote with 99 percent of precincts reporting,” compared to Miller’s 34 percent.

And consider this bizarre twist: Lt. Gov. Craig Campbell told the AP “that since Miller’s name isn’t on the official list of write-in candidates, any ballots with ‘Joe Miller’ written in won’t be credited to” Miller. Campbell changed his tune the next day, and a follow-up AP story reported that write-ins for “Joe Miller” would count for Miller, after all.

The Alaska Senate election is a mess. But nobody ever promised that representative government would be easy. If we want tidy and easy then we should have a king or a dictator.

Even before the Seventeenth Amendment passed in 1913, changing election of Senators from state legislatures to the popular vote, the legislatures themselves were popularly elected. Members of the House always faced a popular vote.

Some years ago your senior author Linn lived in New Mexico and got involved in another important write-in campaign, this one for Joe Skeen.

Politics in New Mexico makes politics in Colorado and Alaska seem calm and peaceful. Many voters live on reservations, which largely retain sovereignty. Spanish culture remains strong in parts of New Mexico, and in the 1960s and 70s land grants protected by old U.S treaty fell under heated dispute.

Quite a few people from New York and New Jersey settled towns like Rio Rancho. Occasionally their friends and relatives from back home would ask if they used American money or had to exchange dollars for pesos. Those friends would have to be reminded that parts of New Mexico were culturally vibrant long before the British colonies took off.

Party lines often didn’t mean as much in New Mexico. Republicans often worked with Democrats. A lot of people were what Coloradans might think of as “Texas Democrats” — fiscal conservatives who would make some of the local Mesa County “conservatives” seem like flaming liberals.

Wikipedia offers a good summary of Skeen’s write-in campaign for Congress: “Throughout the 1970s, five-term Democratic Congressman Harold Runnels had been so popular that the GOP didn’t even put up a candidate against him in 1978 or 1980. Then, on August 5, 1980, Runnels died of cancer at the age of fifty-six. The state attorney general, a Democrat, announced that the Democrats could replace Runnels on the ballot but that it was too late for the Republicans to [add a candidate]. Republicans were outraged and rallied behind a write-in effort by Skeen, while the Democrats selected Governor Bruce King’s nephew, David King, over Runnels’ widow, Dorothy Runnels.”

Those critical of Governor King called his political move “nephewism” rather than nepotism.

After Dorothy Runnels launched her own write-in campaign, Skeen won the split vote with 38 percent of the total. Wikipedia adds that Skeen, buoyed by the popularity of Ronald Reagan, “was only the third person in U.S. history to be elected to Congress as a write-in candidate.” He served until announcing his retirement in 2002.

For those who fought for Skeen, this was a joyous and triumphant victory.

This year Colorado ballots offered the ability to write in candidates only for a few offices, and then only among officially recognized candidates. In many races voters could “choose” only a single candidate.

Statute 1-4-1102 specifies that an affidavit for a write-in candidate must be filed “by the close of business on the seventieth day before any other election” besides a primary. That’s ridiculous. What if an illness, death, or major scandal occurs within that time? What if, for instance, the Republican candidate for governor had imploded closer to the election?

On November 3, Lynn Bartels of the Denver Post reported that Kathleen Curry, a write-in candidate for House District 61 who previously held the seat as a Democrat, challenged a preliminary finding that she came in second. She claimed that voters who wrote in her name but didn’t check the nearby box would put her ahead if properly counted.

If we’re serious about letting voters select the candidates of their choice, we will let people easily file as write-in candidates right up to the election. And we will ensure those votes are properly tallied, even if it takes longer. Do we want speed and convenience, or do we want to let people vote their conscience?

Anonymous, Verifiable Voting

Very often I agree with Vincent Carroll. But not this time. In his December 27 column for the Rocky Mountain News, Carroll claims that paper ballots are “18th century technology.” He points out that mail ballots are also problematic, as “[e]very unwanted ballot is an invitation to attempted fraud.” He concludes the section:

The point is not that mail balloting or paper ballots are rife with fraud and error (although mail balloting is clearly the sloppiest system of all), but that we should weigh relative risks before stampeding out of the electronic arena. After all, if I can buy stock electronically without worry, why should I still have to use a pencil on Election Day?

I agree with Carroll’s criticisms of voting by mail, which is why I oppose the practice. Of course, nobody is arguing that we must “use a pencil” to vote; that’s just a straw man. The issue is whether the vote should be recorded and counted purely digitally, or whether the vote should be recorded and counted via physical records, such as printed or punched paper. (See my earlier post.)

Following is a quick e-mail that I sent to Carroll:

“After all, if I can buy stock electronically without worry, why should I still have to use a pencil on Election Day?”

The answer to your question is simple. When you buy stock electronically, you can verify the transaction online. You can verify the transaction by phone and by regular mail, if you need to. If somebody steals your stock, you will become aware of this, and you will be prompted to take corrective action.

When I vote electronically with no paper record, I have absolutely no way to know whether my vote was counted at all. Nor do I have any way to know whether my vote was counted as I cast it. What if one or more machines malfunctioned? What if somebody tampered with one or more machines? It’s quite possible that absolutely no physical evidence would exist regarding such problems.

True, paper ballots can be “lost,” miscounted, or altered. But at least there’s a much better chance that such problems will yield physical evidence. Assuming that multiple parties always watch the paper ballots, it’s much harder for a single person to change or destroy some of them.

I’m all for modern, mechanical, computerized voting systems. But I also want reliable, verifiable results. And that requires a physical record.

The problem is that voting must be anonymous. Sure, if each voter could cast a digital vote that recorded the identity of the voter, these records could be verified. But nobody doubts the logic behind anonymous voting: it is required to prevent coercion. We don’t want union bosses, gang leaders, employers, politicians, or bureaucrats to know how people voted. Yet voting totals must be made public. (On the other hand, Vincent Carroll’s stock transactions need not be made public.)

So how do we verify vote totals when each vote must be anonymous? The only way to do it is to allow voters to generate a physical record. It’s totally fine for a computer to assist in the process. But, ultimately, the output must be something more tangible than a magnetized blip of a hard drive (or the equivalent). (Has Carroll never suffered a computer error, failure, or virus?) Then, the physical records must be carefully monitored by multiple parties, transported to counting centers (again while monitored and protected), and then counted. Machines can do the counting, so long as the process and the results may be verified by human beings. These physical records must be accessible to legitimately interested parties, subject to proper security.

No form of voting is absolutely fail-safe. But a system of protected and monitored physical records is difficult to abuse, and the magnitude of abuse is bound to be minimized. A purely digital system, on the other hand, allows no method of verifying the vote. Such a system will prove a constant temptation for those clever with machines. Abuse of such a system is virtually guaranteed. And we are unlikely even to learn of abuse when it happens.