After I posted my discussion of the accuracy of Washington state's manual recount, I saw two issues raised in comments here that deserve follow ups.
First, some wondered about the existence of systems that would print paper ballots for voters. As far as I know, no such systems exist, but I sketched some ideas for that kind of system in this post. (I doubt that I was being original, but I haven't seen similar proposals elsewhere.) The basic idea is to use machines to create paper ballots that can be read both by voters, and by separate vote counting machines with very high accuracy. And I should add that I received some interesting suggestions from a partisan Democrat for improving on my sketch. I am still thinking about the problem and would welcome more comments.
Second, I saw the claim, more than once, that an MIT/Caltech study had shown, contrary to what we would expect and most election officials believe, that hand counting of votes is more accurate than machine counting. Here's the key quote from the study:
Have we made progress? Do machine counts improve on hand counts? At least in the comparison of optical scanning and paper, the answer is yes. Historically, there is about a 1 percent difference between initial counts and recounts when ballots are tabulated by hand. The discrepancy between initial counts and recounts falls to about .5 percent with the optically scanned ballots.
That's still way too inaccurate in my opinion, but it seems clear which method is better. (Those who want to look at the study should also look at my criticism of one of their metrics, often misused by Democrats in the arguments over this last election.)
Cross posted at Jim Miller on Politics.
Posted by Jim Miller at January 14, 2005
07:51 AM | Email This
And let the world know: KVI said today X-TINE is trying to keep the people of Washington from finding out who are the felons are across the state that voted.Throwing up legal roadblocks to this very important research. This woman looks more and more like a tyrant every day. Make sure everyone knows what she's doing. She'll be even more hated.
It makes me realize that if they are trying to do this, republicans must have a very good case! But stop Queen Christine! I'm going to tell everyone I can what she is doing. The people have to know.
Posted by: Michele on January 14, 2005 08:27 AMWhen I don't know enough about any candidate for a position, I don't vote. I believe that is more responsible than picking the first one in the list, or some other method of formula voting.
Maybe I should begin writing in a candidate for those positions to make my selection crystal clear.
By separating 'ballot preparation' and 'vote counting' there's no receipt removed from the premises, and there's a physical piece of paper that anyone contemplating fraud needs to dispose of _and_ replace with their own. (The replacement part could be made insanely difficult if the 'voting machine' stamped a guaranteed empty area "Rec'd Machine #123 Precinct #145 Time: 4:23 PM Ballot #1243 of day") Leave space for each count and recount.
Posted by: Al on January 14, 2005 09:33 AMYES!!!
The benefits of accuracy and verifiability together! This reduces the temptation to "improve" the numbers, either mechanically (having a couple of extreme party "loyalists" in a back room churning out extra ballots) or by programming (having the folks that designed the system switch a few votes around.)
Whether or not either of those things ever actually happened in the past (another debate entirely,) everyone can agree that they should not happen, and that by making it more difficult to do (and easier to detect) we improve the system.
I have always believed that the part of the prayer "lead us not into temptation" includes an obligation not to lead others into temptation. This is why, even though I trust my neighbors, I still lock my door when I leave for the day.
This type of system reduces the temptation to cheat, both because it would be easier to detect, and because both sides could have some confidence that the other side was playing it straight. Some folks will never be convinced of that, but it would help.
Think of it. No "hanging chads," no "voter intent," no fellow with a pocket protector (and no social life) switching every tenth vote, no teenager in the back room churning out extra ballots. What's not to like?
No system will ever be completely resistant to deliberate cheating, nor will it be entirely resistant to idiots and fools (whenever you make something foolproof, someone else invents a better fool,) but this is certainly a big step in the right direction.
As to the cost, well, let's compare that to the cost of a new election this time around.
The only people that would have a difficult time endorsing this type of change are those that want the ability to cheat, one way or the other.
Posted by: John Barelli on January 14, 2005 09:42 AMBut this also brings to mind that a person's vote is supposed to be private. Under no circumstances are you supposed to be able to find out how a person voted. That could bring the possibility of intimidation by an organization based on how a person votes. Plus the security and encryption that would be required on those voting records that undoubtedly SOMEONE would hack into for whatever reasons...
I'm not saying this method can't be done, but I am saying that there is more to this than just making it easier to vote accurately. You also have to weigh in the security of the vote and how to keep others from abusing the system. Otherwise you would get what we have here all over again.
Posted by: Netraven on January 14, 2005 10:36 AMThe paper I am quoting from covers New Hampshire recounts. Sorry for the mistake.
Posted by: Jim Miller on January 14, 2005 10:50 AMNetraven commented:
"One minor problem is what if someone pays attention to the ballot and recreates it on his home computer? I suppose you can use the "Lotto" Method that has it's own special paper with serial numbers on it and each one is bar-coded with a serial number that will recall the voting record."
This would be solvable using identification by date/time/machine and checksum, but without personally identifying information. Duplicates would stand out like a sore thumb, and would take more to copy than just "observing" the ballot. Optical scanners get rather picky that the spots are in exactly the right place, although special paper would be a good safeguard as well.
"But this also brings to mind that a person's vote is supposed to be private"
Solved the same way. Auditors could find out which machine was used (and when,) but not which voter used it. Mysterious votes cast before and after the polls close would also stand out and be easily tracked (and eliminated.)
Accuracy, privacy and verifiability, along with a built-in backup for those cases where the counting machines have problems. This addresses the concerns of both sides in this debate.
Yes, with enough effort and enough deliberately dishonest precinct workers, a way could be found to come up with faked votes. But it would be much harder, require many more people in the conspiracy and still be much more likely to be detected. A big step in the right direction.
Posted by: John Barelli on January 14, 2005 11:25 AMAnd it would be wrong to conclude from this study that a hand recount of paper ballots is less accurate than a machine recount of optical scan ballots. Since the results of the recount are considered the most accurate results, all this study tells us is that the preliminary hand count is less accurate than the preliminary optical scan count.
The study's results are based on the assumption that more care is taken in conducting a recount, and thus the recount results are more accurate.
I have looked for a study that examines hand recounts of optical scan ballots, and have yet to find one.
Posted by: David Goldstein on January 14, 2005 11:28 AMSome foolish people have proposed an internet voting system for absentee voters, but that easily raises the correct fear of hackers stealing votes. No system that could be used by the general public could ever be made secure, in my opinion.
Interestingly, the one state where nobody to my knowledge has suggested anything remotely wrong with the electoral process is, of course, the neighboring state of Oregon, where you have 100% vote by mail. The system appears to work as close to perfect as a voting system can work, with little or no problems.
So maybe Washington should look to emulate its neighbor and forget about all these fancy and expensive technological systems that would all be inherently flawed.
If you must go to that, however, I would also suggest a look at Nevada. For the 2004 election it used 100% touch-screen voting and had a printed receipt available for every single voter. The election for president was very close there but the system was very smooth and pretty close to error-free with no allegations of serious errors or problems.
Just a couple of thoughts of how other states are approaching an effort to accomplish error-free voting.
Posted by: Nelson on January 14, 2005 11:31 AMThe "ballot preparation machine" is just making a neatly printed sheet of paper (explained previously.) YES, you probably could make something identical at home... so what? I can make current King County Ballots at home too, what prevents me from casting them? There isn't half the security features of a one dollar bill in these things.
The security is in the pollworkers checking the voter registration rolls. This is precisely the step we already know there's problems with - 1800 "extra" ballots COULD (not alleging _did_, just ONE possibility) have been smuggled in after being 'counterfeited'.
My choice of method would have the second machine sitting on the pollworker's table. Shoving a sheet in is fast relative to filling the whole ballot out. So you have to pass muster (with filled out ballot in hand) to shove your sheet into the second machine. Which puts a timestamp, ballot number, machine number, and precinct number on the ballot _then_. Add special ink and a security hash number or whatnot as desired.
So the poll workers are watching as you actually vote. Sticking a second sheet in is fraud - just like today. Walking out with your ballot is _not_ fraud... and it hasn't been counted as a vote either. (Nothing preventing you from going in and printing up a completely different 'prospective ballot'. But you only get to stick one in the _second_ box.)
Absentees are more problematic - as they always will be. One step would be making the inner 'security envelope' printed in a more counterfeit-inhibiting method itself, then making ballots not placed in correct envelope invalid (which they already are). Additionally, the outer envelope needs your _printed_ name. Either 'voter must print name' or KC provides the envelope pre-printed "Name: John C. Voter Signature:_____ Invalid if not signed yada yada yada"
They have a live Web demo of their proposed software up on their Web site. It looks good to me, and a few minutes' thought didn't come up with any ways to defeat their system. (I thought of a few ways one could *attempt* to defeat it, but each one was easy to counter.) It's definitely worth a look.
Posted by: Robin Munn on January 14, 2005 12:02 PMNelson wrote:
Interestingly, the one state where nobody to my knowledge has suggested anything remotely wrong with the electoral process is, of course, the neighboring state of Oregon, where you have 100% vote by mail. The system appears to work as close to perfect as a voting system can work, with little or no problems.
Well, we actually don't know that there were no problems in Oregon. Had the Governor's race here been won by a few thousand votes, only a very few would even be talking about this now.
I'm of the impression (not shared by many in my own party) that voting is something that is important enough to make me roust my lazy behind from my easy chair in front of the TV and drive the mile-and-a-half to my polling place.
No doubt that the handling of absentee ballots needs to be improved as well. Requiring a postmark prior to the date of the election seems a very easy thing to do.
Side note - I'm retired Navy, and spent many years onboard a number of ships. Please do not trot out the story about the ship that could not get its mail postmarked. If true, then the Postal Clerk involved should have been busted for incompetence and dereliction of duty, then sent to Deck Department and put to chipping paint under close supervision. Postmarks on ballots are important, and every Postal Clerk in the military knows it.
If you must go to that, however, I would also suggest a look at Nevada. For the 2004 election it used 100% touch-screen voting and had a printed receipt available for every single voter
Unfortunately, this does not allow for any sort of manual recount if problems with the machines arise. It would be entirely unworkable to say "we think there may be a problem with the programming here, could everyone please bring in your reciepts?"
Anecdotally, I would also show the example of the New Mexico precinct (from VotersUnite.org):
San Juan County, New Mexico. November, 2004.
1,843 election-day phantom votes reported in
Precinct 51.
With only 318 people casting election day
ballots on the Danaher Shouptronic electronic
voting machines in the Precinct 51, Fran J.
Hanhardt, Republican incumbent running for
County Clerk, received 2,079 votes, while
Democratic challenger Glojean B. Todacheene
received 82.
However, Ms. Hanhardt didn't need the 1,843
phantom votes to win. She won by a strong margin
of 16,484 votes.
Obviously, this was simply a computer glitch. Not even the most partisan Democrats (and certainly not your friendly token Democrat here) are trying to claim that this was deliberate.
Still, you can see the problem. This particular problem didn't decide any races, but in some races, it could easily have thrown the results into question.
Posted by: John Barelli on January 14, 2005 12:11 PMI agree with most of what you said. I am not aware of the NM incident, but I do know of a similar one in North Carolina where a voting machine simply lost more than 4,000 votes, as well as another one in Ohio which gave Bush an extra 4,000+ votes. However, as I understand the Nevada system, a recount is indeed possible, since when the voter gets the receipt, he or she is supposed to look at it and verify that it indeed contains the votes that the voter desired. That receipt is then deposited in a ballot box in the polling place to offer the opportunity to have an alternative count independent of the electronic machine tally in case there is a question about the machine tallies.
That's what the Democratic Party all over the country is urging as the appropriate "paper trail" for touch-screen voting. California will put the system in place in the '06 election in every county that wants to use touch screens.
The League of Women Voters, after some hesitation, has also now endorsed that paper trail system.
Posted by: Nelson on January 14, 2005 12:23 PMThat's what the Democratic Party all over the country is urging as the appropriate "paper trail" for touch-screen voting. California will put the system in place in the '06 election in every county that wants to use touch screens.
Any method that provides for voter verification of the electronic machine, along with a paper-trail for manual counting when necessary would go a long way towards bringing back some confidence in the vote counting. I happen to like the "machine produced paper ballot" method described here, but countable paper receipts, kept by the precincts for verification purposes would work too.
There are a couple of bills going through Congress that would require this type of safeguard on electronic voting machines. They are having a very hard time getting out of committee, and are probably dead for this session. (VerifiedVoting.org has more information on s 1980 and hr 2437.)
I really don't think that this is a Republican/Democrat thing. This is more of a "let's get this mess cleaned up" thing. I'm a proud Democrat, and believe that honest elections help my party. Most Republicans believe that honest elections help their party, too (and they're right.) The vast majority in both parties would prefer to lose honestly than to win by cheating. (Of course, winning honestly is much better.)
Unfortunately, there are some (in both parties) that will never sign off on any idea that comes from the "other side." This seems to be the big holdup on the two bills stuck in committee. So, hopefully the Republicans will make similar proposals (and I'll be writing to our Senators and my Representative to urge them to co-sponsor them.) This issue is bigger than party loyalty.
The only way somebody could vote twice would be to present themselves at DMV with enough stolen or phony identification documents, such as a stolen or forged birth certificate, to create a whole new persona in the database. That's a lot of work and stress just to get an extra vote.
Undervotes and overvotes would be moot. That is, the ballot is handled the way it's presented by the voter. If the voter makes a mistake in his vote for any particular office by voting for more than one candidate or failing to vote for that office, that should be final with no application of human "divination" used to sort it out. If a voter overvotes and realizes it, he may get a new ballot by submitting the spoiled ballot into the on-site database scanner which will read the barcode and spit out a fresh ballot when the voter scans his license again - no new ballot without the license, obviously. The machine could be designed with a built-in shredder.
Now all we need is somebody to invent the machine and design the database!
Posted by: KittyBurglar on January 14, 2005 02:01 PMThe alleged illegal ballots (excess votes, felons, dead people) amount to several thousands (not clear how many, and not clear whether they would have changed the election result). So even if you throw those votes in, you're still probably below the 1% error rate. Granted this is too high. So we need to change the technology and system as you have been discussing. Revoting will presumably have the same error rates, unless procedures are changed, and I think it is well worth putting in the effort to do that.
Posted by: Martin on January 14, 2005 03:41 PMThere's an obvious major problem with your proposal. You're automatically disenfranchising every legitimate voter who, by choice or necessity, does not have a driver's license. This would include the disabled, the elderly, the infirm, the poor who don't drive and anyone who simply decides they don't want or need a driver's license. How would you handle that?
Further, there's one thing that everyone is missing in this discussion, and that's costs. One of the main reasons why voting is never as accurate as, say, bank accounts, or even DMV databases, or ATM machines, etc., is that virtually all other activities occur on a continuing basis throughout an entire year and have individuals working on them on a continuing and professional basis.
Elections, however, occur only on 1 day every two years and have largely volunteer help working on that day. So while voting is critical in our democract, and accurate vote counting is an ideal to strive for, the fact that systems will get only one day's use every two years and volunteer operatives (even ones who are very dedicated and committed to excellence) may easily forget from one election to the next who all the systems work.
Since they are usually working with inferior equipment because govts are loathe to spend money on equipment that gets such limited single-purpose use, and because they are mostly citizen-volunteers who typically undergo 1 hour of training, the more complex you make the system, the more likely you will have major errors and omissions.
Think simple, efficient and inexpensive and then you'll have a great voting system. And then when you realize that those 3 goals are often mutually exclusive, you would realize why we have the systems in place that we have and why the American democracy will always be a beautiful, but flawed, system.
We've lived with these flaws effectively for 228 years as a country. No problem.
Posted by: Nelson on January 14, 2005 03:50 PMI have no particular objections to the use of electronic equipment to produce machine-readable ballots except cost, which I see as being substantial. Optical-scan systems are good because they allow many inexpensive voting stations to be used in a precinct with one scanner. I would, however, add a small but significant change: for every candidate position, there should be a "NO CANDIDATE SELECTED" option (in "vote for N" offices, there would be N options for "NO CANDIDATE SELECTED", "ONLY ONE CANDIDATE SELECTED", etc.) For a ballot to be accepted by the tabulator, it must be perfectly filled in. To help save voters time, there could be a "validator" provided which would automatically fill in the "NO CANDIDATE SELECTED" ovals for unvoted races and hand the ballot back to the voter (who could confirm that it had done everything correctly). Under no circumstances should the tabulator accept any ballot which is not perfect and complete, or on which any oval is between 25% and 75% filled in, or on which any stray marks appear.
For absentee ballots, I would add two new fields, one of which would explicitly state the total number of candidates selected, and the other of which would explicitly note the number of corrections made. Voters would be instructed to complete these fields last, but that completion was required. This would allow a voter to make corrections on his ballot prior to completing the "number of corrections" field, but ensure that nobody else could do so after the fact.
Finally, I would explicitly provide in law that the proper means for doing a "hand recount" is to have the machine mechanically sort ballots into piles based upon the vote read for a particular office, and have the ballots in each pile examined to determine if there were any that were out of place. Possession of any sort of black pen in the room with the ballots would be a criminal offense, and signs would be conspiciously posted to that effect. Any ballots which were found to be wrongly sorted should be marked with control numbers using a purple pen, placed in special piles, and tracked in detailed logs.
These changes should not significantly increase any of the costs associated with the election beyond the expense of one validator per precinct (a much smaller cost than requiring many ballot-printing devices per precinct) but would provide a level of security far beyond what exists now.
Posted by: supercat on January 14, 2005 04:51 PMCall my cynical, but I personally suspect that RINO candidates rely upon Democrat voter fraud even more than Democrats do. After all, Democrats can respond to a "rightward" shift in electoral patterns (such as might occur with Democrat fraud eliminated) by running to the right. RINO candidates, however, do not have the same luxury. The only reason anyone bothers with them is a perception that more conservative candidates "aren't electable". If it became clear that conservative candidates were electable, nobody would have to bother with faux-conservative RINOs.
Posted by: supercat on January 14, 2005 04:57 PMI too find bothersome the fact that "undervotes" are not treated as an explicit and deliberate decision, not only because they're considered "errors", but also because an "undervoted" ballot can be altered after the fact to indicate a candidate for whom the voter did not vote.
I would like to see voters required to either indicate explicitly that they did not intend to vote for all offices, or have a machine explicitly mark a ballot to that effect in a way that a voter could inspect and agree to.
Posted by: supercat on January 14, 2005 05:07 PMThere are a few proposals buried out there for an electronically-prepared, human- and machine-readable ballot, but most focus on a readable receipt available for audits and recounts, with the actual ballot recorded directly and electronically.
Included in the initiatives are some open-source projects. Unfortunately, I don't have my bookmarks handy so I can't include some of my favourites (e.g. federal standards for voting systems).
Posted by: Paul on January 14, 2005 05:20 PMOne difficulty I see with this approach is that if a person is allowed to handle the physical artifact associated with a vote after the vote is counted, it becomes impossible to insist upon perfect reconciliation between votes and "receipts".
If one wanted to have a system in which receipts were held in a 'captive window' before being fed automatically into the ballot box, that would be doable. I would expect, however, that requiring one such machine per simultaneous voter would be more expensive than having one validator and one scanner for every 10-20 simultaneous voters.
Posted by: supercat on January 14, 2005 05:35 PMThis is the single fact which most people miss, so it bears repeating: using a scanner-based system requires (perhaps) one machine per voting location; other systems (e.g. touch-screen) require one machine per simultaneous voter. The scanner systems are thus much cheaper to use in an election.
Posted by: Paul on January 15, 2005 02:06 AMhttp://www.cev.ie/htm/report/download_first.htm
Posted by: Paul on January 15, 2005 02:10 AMEveryone should just walk into a Chili's Restaurant, and watch what happens when the waiters enter an order into their system. They log in with either a magnetic slide-card or a PIN, use the touch screen to generate the order. Said order is BOTH printed out at the order station, and printed at the cook's station in the kitchen (as well as going up onto a display board). At the end of the meal, TWO copies of your meal details are printed up - one for the waiter, one for the customer. Additional copies are printed if using credit card.
All this talk of the "technology not being available" is out-and-out bunk. The problem is that they are valiantly attempting to make the simple extraordinarily complex. If Chili's can pull something so similar like this off, why can't the election boards???
Posted by: JD on January 16, 2005 06:28 PM