In the USA, there is a current debate on the security and verifiability as regards voting machines, audit trails and verifiability. The current trend is towards electrnic voting machines, from multiple manufacturers, with Diebold being the most prominent name.
These electronic voting machines are pushed as a replacement for the now-ageing mechanical voting machines (effectively punching a card that can later be tabulated by a tabulation machine). With the mechanical machines, the resulting punch cards work as an audit trail that can be independently counted and verified. The current crop of electronic voting machines have much les sin the way of an audit trail, there is nothing outside the machine that can be verified by the voter as correspoinding to the vote they cast and nothing that can be independently counted afterwards. Pair this with several well-known weaknesses of the system (the ability to add code to the machines that can change the vote tally after the fact without affecting the audit trail, common locks with keys that can be reproduced by cutting a key depicted on a Diebold web page and many more).
In Sweden, the voting process does not suffer from these kinds of weakness and is a possible model for future US elections. It is not clear that the system would scale to the required number of voters, but I suspect it would. In the Swedish 2006 election, there were a total of 6 892 009 people eligible to vote, 6 766 697 of whom were resident in Sweden at the time. Of these, slightly over 5.7 million did vote. I do not have a break-down between postal votes (absentee ballots) and normal ballots.
The Swedish election process is suitable for multiple parallel "races" (essentially independent elections run on the same day) and has a functioning model for postal voting (though this requires a setup where the independence of the vote cast can be verified).
The basis of the voting process is designed around voting in a polling station, the postal vote is similar, though it will be described in a later section.
The basis of the system is a paper ballot, marked clearly on the page itself with the race and the specific vote to be cast (these come pre-printed for most, if not all, eligible votes). There are also "blank ballots", that are only marked with the specific race and otherwise unmarked, they can be filled in by the voter.
The ballot marking is done in text on the paper, stating something along the lines of "Municipal election", "State election", "Regional election" and the like. All ballots are also marked along all edges, half-way between the corners, with a symbol signifying the specific race. After the ballot(s) have been selected, they are put into special envelopes, with one perforated long edge, a flap and glue strip on the opposite edge and small circular cut-outs half-way along the edges, so you can see the symbol signifying the race of the enclosed ballot.
In addition to this, there is a (paper) list of all voters that have been assigned to a given polling station (if you cannot vote at your assigned polling station on election day, you can do a postal vote in advance). On handing the ballots in, the voter is marked on the list, so that one can adhere to the "one person, one vote" principle.
The polling station workers then receive all the ballots cast by a given person, mark them off against the list and inserts the ballots into race-specific ballot boxes, through a slit cut in the lid. These ballot boxes are sealed with a crimped seal of some sort. When filled, supplemental ballot boxes can be brought into function.
Once the polling station has been closed, it is time to count the votes. This is done by manual count. A group of people are tasked with opening the ballot envelopes and check that two things hold; there is only one ballot paper inside the envelope (this can be done prior to opening) and that the ballot is for the right race (again, this can be verified before opening). Anything that seems to not meet these conditions are set aside, for later verification. The ballot envelopes are then opened and distributed to a counting station responsible for a given vote. They are then counted and stacked in groups of 100 and passed on for re-counting.
During the counting, any ballot that is deemed "defaced" or "not valid" is again put aside for later verification. All parties in the race can send representatives to monitor the vote counting and there is also a provision for independent parties to turn up.
All this counting is done on a per-polling station basis. After the votes have been collected, a phone call is made to the next higher instance (for a polling station, this is an electoral district), so they can upate a running tally for the electoral district. The count is also written down on an official vote registry and all ballots are packed up. The written tally, the marked voting registry and the actual ballots are then couriered to the electoral district counting office, where they can be stored for fture verification.
This system is also semi-friendly to postal votes. The way the process works in Sweden is tha you cast a postal vote if you are unsure you can attend youre polling station in person on election day. This can be done for several weeks in advance, at embassies, consulates and (in Sweden) post offices. The process is similar to how itr works at the polling station. You stick a ballot per race inside a voting envelope, then take these up to be registered. The main difference is that these are then stuck inside a larger envelope, marked with sufficient identifying information to tie this bundle of votes to the voter and marked with the time they were handed over.
The main advantage of doing it this way is that you can then, if you suddenlt find that you are able and that you want to change your vote, go to the polling station and vote again.
Before the postal votes are even opened, the identifying information is compared to that of your "home polling station" (as previously stated, you can only vote in one specific polling station on election day) and any postal votes that can be found by a person that voted in a polling station are immediately destroyed.
This, of course, means that postal votes are counted only after all of the polling stations have closed. In fact, they are usually counted up to a week after the main election. Not as much of a problem as one would imagine, since it's usually a very small proportion of the votes cast.
A count (discounting the postal votes) of all votes, essentially accurate, is usually available at 23:00 on election day. Polling station close at 17:00. There is usually some recounting needed of individual districts and this takes a bit more time, as what is and isn't a valid ballot is haggled back and forth.
Note that the counting process is (essentially) very parallel in nature. Each polling station caters for a couple of thousand voters. Each district is composed of tens to hundreds of polling stations and if needed several more layers of indirection can be created, so as to scale upwards to tens, maybe hundreds of millions of voters.
One concern is the sheer cost of producing voting material. With pre-printed ballots in sufficient numbers to cover all possible electoral compositions, the number of ballots scale as O(Nv * Nr * Nc) (where Nv is the number of potential voters, Nr is the number of races and Nc is the number of candidates in any race (well, the average number of canbdidates over all races, but that is just being picky)).
This can be improved to O(Nv * Nr) if you either postulate having only blank ballots, for each voter to fill in OR ballots with all candidates pre-printed, so you can just tick off the candidate you are interested in voting for.
With the current model of postal votes (absentee ballots) in Sweden, there is a risk of breaching voting confidentiality, in that you have a nice package, conntaining all votes plus identifiable information. This is non-trivial to solve, since the intent is that you should be able to discard votes from people who both cast an absentee ballot and one in person down at the polling station.
There is a small risk of vote-buying in the way the postal votes work, since there's usually less surveillance of that sort of thing in the ad-hoc locations used for postal voting. This is largely balanced by the abilit yto change your vote by voting at the polling station. It can also be improved by only counting the last postal vote (not sure if that is the case in Sweden at the moment, though).
The voting process itself doesn't pose any uncorrectable problems, there is a paper trail going back to the polling station (the connection between "voter" and "ballot cast" is intentionally broken, so a sufficient amount of skullduggery can be used to subvert the ballot boxes on a polling station by polling station basis; the amount of concerned, non-aligned observers makes this unlikely).
There is a small risk of misrepresenting or subverting the polling sttaion count on its way up the chain to the central-most collection point. This is, in part, balanced by the couriering of ballot boxes with an independently verifiable summary sheet.
On the whole, it is a system that is at least as scalable and at least as secure as the current American practice of using digital vote recorders. It is, in all probability, more secure. However, it means less income for voting-machine vendors, even if it provides an opportunity for specialist election paperware manufacturers.
This is one of Ingvar's essays
By: Matt
2009-12-02 15:11
Is there a similar style voting system used for the rest of Scandinavia? To my knowledge, most countries use a paper ballot voting system. The United States primarily uses electronic or mechanical voting machines, some without paper trails. This lack of paper has, on numerous occasions, resulted in electoral problems that affect vote outcomes. Any comments welcomed.