How to Count Thousands of Paper Ballots by Hand

How to Count Thousands of Paper Ballots by Hand

by John Washburn

August 17, 2005

Introduction

A growing number of election integrity advocates and activists are drawn inexorably to the conclusion the best system available is paper ballots marked by the elector and counted by hand. This system is transparent, understandable, reliable, and provides for recounting in the event of an election dispute.

The single greatest objection presented by vendors of election machinery is that hand counting thousands of ballots would be prohibitively time consuming and prohibitively expensive. Neither claim withstands scrutiny.

This paper describes how to efficiently count ballots by hand with the aid of a good counting scale.

snip

http://www.washburnresearch.org/HandCountingPaperBallots.htm

Counting using a tally sheet would take 117 man-hours.
Counting using a counting scale would take 61.5 man-hours.

Unfortunately, I'm afraid it probably won't fly. I can hear the echo chamber now: "Weighing our votes!!???! Now I've heard everything!"

It can be highly accurate, especially with a digital scale. I'd go along with it.

As someone who receives small parts that are weighed by the manufacturer, they are not always on target. Most manufacturers will give you a handful more on top of the weight to make up for the error. Plus I can now see the tinfoil hat people telling us that Mettler Toledo was big republican donators and that their scales were rigged to weigh republican votes.

If you really want a machine, get the vegas people to use their money counters. make the ballot in the shape of a dollar bill and use money counters.

And so bloody what if they don't?

Better the count take a week and be truthful.

:thumbsup:

the numbers off? Would they do it through paper weight variances? 75lb card stock vs 80lb card stock, etc. Who knows.

Many countries, including Canada, hand count every ballot. That's still the ideal method, IMO.

Canada doesn't have a zillion different races on each ballot, nor do we hold so many elections on the same day.

AS a Canadian, I used to argue that we can do it, so why can't the US, and eventually I realized that it's the volume of ballot-races that would slow things down. Our federal elections are usually one race on one ballot and are counted in 45 minutes. Your elections include many, many different races, which would make things very slow - not that THAT'S a bad thing.

HG

A number of hours -- or even days -- of counting vs eight years of the BushCo reign of terror.
Which would you choose? I don't know if this country will ever recover. (To say nothing of Iraq)

Other options suggested included the hand counting of major races and issues only.

different races for different districts; i.e. different congressional districts, etc. Some cross over into a number of precints while others cross over into other precincts. Then there are the ballot measures, which can be twenty or thirty. There may be well over fifty different items on a single ballot, and combine this with up to 170 ballot styles in a single county, and it gets very complex. We'd have to go to an entirely different system of balloting.

How about asking John Washburn what he has to say to that?

it wouldn't matter how long the hand counts took. We would still know who won election night.

I just don't see how we could ever have exit polls that we know to be that reliable. It's surely impossible in an election that comes down to several hundred votes out of millions. But for instance in Ohio in 2004, even if you assume that the exit poll was unbiased, there just wasn't enough information to call the state. And I can't think why anyone would ever bet the ranch (never mind the future of the country) that any particular exit poll was unbiased. We could change a lot of things about exit polls, but I don't know how we could perfect them.

1. Are you saying there is no way to improve our exit polling?

2. Assuming we can't improve our exit polling are you saying we shouldn't hand count?

They said their last election took until 4:30 am, to hand count/announce the results. Bush wasnt decalred the winner-- until-- what? 3:30 am.

Hmmmmmm

Ever see how long a California ballot can be? Ever see an Illinois ballot? The hand count is orders of magnitude larger for those states. You will never get all the states to have separate presidential elections. The FED can not enforce that. To really reform elections, you have to reform the entire process. You have to reform the requirements for office. You have to make each person elected more accountable. One idea is to make each election cycle a recall election. This way the issue is whether an elected official is doing the job. This way you do not have to use the lesser of two evils. This way the elected official has to defend him/her self from both parties at once. If a recall is defeated, then we do not have the wasteful money spend on other candidates. If the recall is successful, then the candidates only have a month to run thus eliminating the wasteful spending. It also requires a candidate to get out their message fast. They do not have to be negative as much because they have to sell themselves not beat the incumbent.

Speaking for myself, no, I have never seen how long a California ballot can be, although I know they can have more propositions than most mere mortals can begin to think about, never mind elected offices. But can you point us to a California or Illinois ballot? Or a specific count of ballot positions?

I'd like to get to the point where if someone presents a fairly detailed cost estimate, as in this case, we can kick the tires and see how well it holds up -- not just make general points about ballot complexity. It would be great to get the hand-count boosters and skeptics to the point where we can agree about some of the facts, even if we can't agree about what to do with them.

(EDIT: I should note that in this case, you presented a feasibility argument upthread in #9: that even though the scales are nominally accurate enough to count ballots, in practice this just wouldn't work. That would be my guess, too. Maybe a scale can register 8336 grams plus-or-minus 2, but I'm skeptical that that will reliably equal 1042 ballots at 8 grams a shot. Not that any other method is guaranteed to be absolutely accurate, either, but scales seem awfully finicky to me. So maybe the cost estimate has to be all labor. I don't know whether I zoned out the last time a thread got to that point, or whether the thread itself zoned out.)

Your second point is that there is no way to enforce hand-count paper balloting for presidential elections. Hmmmm. The feds manage to enforce (at least de facto) all sorts of other things. I'll punt on this for now. If hand-count paper ever got popular enough to be viable as federal legislation, it probably wouldn't need that much enforcing.

OK, now, the recall thing (but don't let me hijack the thread): the problem with that is, people don't generally pay that much attention to what their incumbents are doing -- and incumbents generally work hard to be seen doing good things. So, that change could make incumbents even less accountable than if folks at least have a shot at mounting campaigns against them. Not that the current system is any utopia of accountability.

This money will buy a counting scale with a capacity of 50 to 70 pounds and a sensitivity of 0.005 pounds (2 grams). A single 8.5”x11” sheet of 80 pound card-stock weighs 8 grams. A ballot printed on heavier card stock or on a larger size would weigh more. Thus the ballot is well above the readability of the scale and is able to distinguish differences of even a single ballot.

I have a ream of 80 lb stock. I'll check out the variability of the weights and estimate the error when I have time.

A scale that is precise to plus-or-minus two grams isn't necessarily accurate to anything like plus-or-minus two grams. In fact, we have a consumer scale (we use it to measure coffee beans) that is visibly inconsistent, albeit pretty darn good. I'm sure a scale that costs fifteen times as much is more reliable, but I still wouldn't assume that it is accurate to 8 grams out of 10,000 or more.

To test this approach, we would need some practical experiments into the scales' variability. And we would also need to decide how much error we are willing to tolerate.

My hunch is that even if this is objectively a good idea, it will never happen because it Just Seems Like Cheating. But I'm still curious whether it's objectively a good idea.

However, precision is more important than accuracy, as long as you use the same scale, and recalibrate frequently. If the "real" value is 7 grams and not 8 grams, it doesn't matter as long as the stock used to calibrate is the same as the actual ballot stock.

(I don't mean to imply any opinions about the measurement characteristics of $1000 scales, just trying to clarify our points -- since I muddled together two points in my preceding post.)

What I called nominal precision, I should have called resolution. A scale reports to within +/- 2 grams: that is its resolution. If it is "precise," it will repeatedly report that a given stack of paper is the same weight (or I guess for grams I should say mass). If it is "accurate," it will report the actual mass of the stack.

It seems to me that, as you say, precision trumps accuracy as long as any inaccuracy is a fixed percentage (or else perhaps a known function) of true mass. If we are weighing a stack of 250 ballots to determine the average mass of a ballot, it really doesn't matter whether the answer is 1% low or 2% high, as long as we use the same scale to measure the stacks we are counting, and as long as the scale is proportionally consistent.

I would think it would be pretty hard to engineer a proportionally consistent scale -- one that is equally accurate or inaccurate for light, medium, and heavy loads -- but I don't know. Maybe a properly calibrated $1000 scale handles that problem like a champ. I also don't know how repeatable the measurements are. My nominal 2-gram scale (which we bought to track the weight of a cockatiel) seems to have considerable random variance. So I have questions about both accuracy and precision.

--for the number of ballots used to calibrate the scale as you would choose for the number of ballots per batch counted. You would not calibrate with 100 ballot sheets and weigh batches of 1000 actual ballots--if your ballot batches are about 1000, that's how many ballots you'd use to calibrate the scale. In analytical chemistry, we are always careful to avoid the problem of nonlinearity of standard curves over several orders of magnitude by choosing our standard concentration to bracket the expected amounts in the samples. If the sample concentrations are too high or too low, then we rerun with an appropriate standard curve.

The stacks are going to vary widely in size -- some races will be competitive, others not. OTOH we could modify the protocol to split up large stacks, if that would help (and, at the other end of the range, we can hand-count small stacks).

There are empirical questions here that I don't have the answers to. John Washburn (whoever that is) thinks it will work, onthebench (whoever that is) says it won't (see #7), and I don't know. Mettler-Toledo claims typical repeatability of 0.002 lb, which is fabulous if it applies to scales that are being delivered to polling places (and then rechecked, of course). Certainly there is nothing to prevent a jurisdiction from trying it, if it is convinced that the idea is a good one.

Oh, apart from the technology -- Washburn's model posits one counting scale, and fifteen counting pollworkers, for a village with 7000 voters. That isn't how it's going to be in most jurisdictions, if the counting is done at each polling place. So to pitch this model to a county, one would have to work up a different budget. That shift might actually benefit the model, I'm not sure. It's not my model.

There is much more complexity if you are going to combine mail-in with precinct polling places. I've downloaded his spreadsheet and am putting King County WA numbers in it. This will be a really rugged test, as only Los Angeles County counts more votes.

I don't think all mail-in is sweeping the country, at least not yet. Personally I don't really think it should, but that's beside the point. Anyway, I'm curious to see what you learn.

--King County would be the worst case scenario.

then how accurate is the count?

--damaged absentee ballots. In King County WA, duplication of ballots is overseen by representatives of both parties and anybody else who wants to look. Ballots get mangled in the mail, sit on top of other ballots filled out with leaky pens, and sometimes are rejected by the opscan machines for no obvious reason at all. These ballots are all duplicated.

Oregon is a LOT smaller than CA.

(I think when I tried to do this for a particular precinct in Cuyahoga County, Ohio, the average turned out to be less than 2 because a lot of races were uncontested.)

Counting ballots 50 (or more?) ways certainly seems time-consuming to me no matter how one goes about it. That's why I'm not a hand-count purist. But what we seem to have right now is a debate where one side says that a hand count (or in this case a hand sort) is "too hard!" and the other side says "is not!", without agreement on how hard it is. If we can specify how hard it is, then we will probably still disagree on whether it is too hard, but at least we will know something together.

go back to hand counting. He is an honest, ethical hard working guy, very committed to quality control and fair elections. Just doesn't get the issue with the vendors and their ability to program malicious code. If you talk to county election officials, they make some strong and very valid points in favor of machine counting. Things are much more complex than they were thirty years ago.

Perhaps separate ballots for all national races would work, but they'd have to be convinced there is a problem first, or we'd have to get legislation mandating it. In some ways these discussions are moot, because we have to work within the system the way it is set up, like passing legislation at a state level. I don't know what the answer is, but I don't know if we can even get into the discussion with election officials if we insist on hand counting.

I like your approach to this issue. I certainly don't think we can get into the discussion with election officials if we are long on harangue, short on comprehension and research and responsiveness. Sometimes sheer intensity can carry the day, but even then the results generally aren't ideal.

If the hand-count advocates want to marshal the case that hand counting is economical and logistically feasible, I figure they can start by trying to convince people like me (and you). Or they can just go off and see where they get. But if folks want to marshal an argument for hand counting, I'm willing to try to see where the argument leads. It's a good use of a discussion forum.

hand counting. After spending a lot of hours with a number of election officials who knew a lot more about a lot of things than we did (their ignorance is with issues with vendors and how easily hacked and vulnerable the source code it, and the nefarious criminal history of the vendors), and were seasoned professionals with a great deal of experience, we realized we would come across as arrogant assholes if we started dictating to them how they needed to do their jobs. We came to a more moderate position; I still don't know the answer because I truly don't think it's possible to secure computerized vote counting but currently we are pushing rigorous audits and no private corporations programming the software and no software that is not open source.

(All this agreement is very agreeable....)

Groups can often accomplish a lot by mastering a subject that the officials they are working with don't know, while also acknowledging that the officials know things they don't. Not that sweet reason is always enough, but it's generally a useful resource. I suspect folks would have gotten farther with the exit poll stuff if they had learned a bit more first.

I don't know whether it's possible for people to develop the "complementary expertise" to convince election officials, generally, that hand counts are a good idea. (Obviously some think so now, but it's what, something like 3% of the electorate.) It seems unlikely to me, but maybe the advocates know (or could learn) something that neither I nor the officials do. So I keep asking questions! Meanwhile, I think rigorous audits are especially important.

On (1), first of all, it sort of depends on what you mean by "our" exit polling. You probably meant U.S. exit polling, but the phrase raises other questions. Presently the exit polls belong to the media sponsors -- which I personally don't think is a terrible thing, but it has its limits. One pretty simple thing that would likely improve the exit polls is to have two interviewers per precinct instead of one; but the media sponsors aren't likely to pay for that, because it isn't likely to be cost-effective for them. Potentially, a volunteer exit poll could remove that constraint, and might yield more accurate results. What I do say, however, is that there is no way anyone can guarantee that an exit poll will be unbiased, in the statistical sense. That isn't (entirely) in the hands of the people who conduct the exit poll; it depends on the people who participate. If we are marshalling people power for election integrity, I doubt that exit polling is the best use, but I dunno, and it's not up to me anyway.

(There could be a whole separate thread or three on parallel elections, etc.)

(2) There is no reason why we have to know the winner on Election Night. We didn't in 2000, and the nation survived. There are some issues with hand counting, but I don't see that as one of them. Sorry, should've said that the first time around. Personally, I'm pretty much agnostic on hand-counting (except that I think that other methods should be validated by at least a partial hand recount -- the alternatives I've heard don't convince me). So far, I'm not convinced that this particular method is cost-effective -- ballots can get very complicated -- but I'm still curious.

1. Current national exit polling is done by people hired by the big 6 media companies -- you may remember big media -- the same people who steadfastly refuse to report on election problems unless absolutely forced to. So, in order to have decent exit polling, it must be independent of those with a stake in elections, including big media.

2. No, we're not Canada. Canada, to my knowledge, doesn't endorse torture at the highest levels of its government. Doesn't invade countries. Has health insurance for all. Has gay marriage. Has prescription drugs that are so affordable Americans buy them and cross the border with them. Right, we're not Canada. (P.S. Canada DOES have big cities, including Toronto, which is huge.)

3. The question is not, gee, do we have too many voters in the precinct or too many issues on the ballot to count by hand. The question is not, gee, can we have the results to the big media in time for the 10 pm news. The question, folks, is, are we a democracy or not? Will our votes count or not? If we determine that hand counting is the best way to guarantee the least amount of tampering (and I believe it is), then we need to adjust precinct sizes so that they are small, and have more frequent elections if the number of things on the ballot are just too many. Period.

and the only way the neocons can be taken out of power,is to get rid of their election theft counting machines, And the Dems and Repugs are silent on the issue, If the Dems and the Repugs would speak in one voice and just say ITS OVER the people are on to us. I for one will forgive them if they used that corrupt vote counting machine to get in to office, so long as they prevent them machines from counting the vote's in the future.

As far as it being to hard to count ballots by hand, thats the Neocons last stand or (excuse) in order to keep them election theft machines counting the votes. That excuse is BULLSHIT! Plain and simple.

A lot of the top dogs in the Senate, for instance, probably would not be there if not for gigantic contributions from corporations whose businesses they help, AND no doubt years of rigged elections.

Look at how many Senators lept up to sign on to an equivalent of Rush Holt's simple paper verification bill -- look at the ONE, count em ONE Senator, Barbara Boxer, who helped challenge the Ohio electors on Jan. 6, 2005, and you get a clue as to how big this problem is. (ZERO Senators challenged the results in 2000 -- recall the scene in Fahrenheit 911 when the Black Congressional Caucus was BEGGING one Senator, any Senator, to help.)

had asked her not to.