Magnetic cards with printed records are used in transportation systems in Belgium, Japan and probably many other places.

The main advantage of magnetic cards is that they are easy to handle by manual and automatic means.

There exists a patent for a voting system where magnetic cards with printed records are used. That patent suggests use of magnetic records for counting, whereas printed records are to be used for verification (if later required). That patent should not apply to voting systems where election results rely specifically on counting printed records.

As much as I know, magnetic cards with verifiable printed records on themselves were not used in elections yet. Magnetic cards are occasionally used in electoral voting (most notably in Belgium), but convenient verification possibilities are absent. But usual objections to electronic voting remain.

2. The voting procedure

The voting procedure may look very usual.

First, a voter should be authenticated by election officials, and receive a magnetic card from them. One side of the card may contain official logos and instructions for voting. Other side of the card is for printing the vote(s).

Secondly, the voter goes into a voting booth, where a touch screen (or a computer, or a screen with a mouse or a light pen, etc.) and a magnetic card reader/writer are available. The voter makes his choice electronically and confirms them on the screen. The choice is recorded on the magnetic strip of the card, and printed out on the available side of the card. Then the card is ejected; the voter should check the printed form. [If the voter does not agree with the printed form, he should be able to return the card to officials and get a new card. A returned card should be handed back the printed side down, invisible to anyone, in order to comply with privacy terms.]

The most important requirement is that the printed form would be well-readable. Quality and correctness of the magnetic record is desirable chiefly for having a smooth counting process. There is no essential need for strict protection of magnetic records. Magnetic records may even be corrected later without objections.

Lastly (after leaving the booth with desired vote), the voter should insert the card into the voting urn and leave the voting area. The voting urn should be able to detect wrong cards and insertion of stacked cards. To give a reassuring impression, the urn may have the form of a subway toll gate and make a similar sound. For convenience, the cards may be immediately stacked into special containers. As a rule of thumb, no one should be allowed to leave the voting area without submitting a card.

3. The counting procedure

This is the essential new contribution. For simplicity, we assume here that the magnetic cards record votes for one nomination. I present here the main counting procedure that I have in mind. In a following Section, some possible variations are described.

In principle, the count of submitted votes can be done in three steps:

1. Sorting the cards by magnetic records, so that cards with the same vote are stacked together.
   
2. Checking the sorted stacks by visually inspecting printed forms.
   
3. Counting stacks of cards with the same vote.
   

Magnetic cards can be transported between different machines in special containers or container frames, so to minimize human manipulation. It is acceptable to touch or inspect cards in a container frame, but it should be impossible to take any cards out. Most care should be taken that cards would not be spread (accidentally or possibly intentionally) over the floor.

The sorting machine should have at least 3 compartments for sorted cards (at least 2 for major candidates, and 1 for other). The compartments do not have to be of equal capacity. Manufacturers should foresee the possibility of overflow of cards with the same vote [or more generally, in the same compartment]. If there is a compartment with different votes, those votes may be sorted in other run or handled manually.

Checking the sorted votes is most likely the most time consuming part of the process. Suppose that a stack of cards with votes for candidate X is being checked. It is most convenient if the printed name of X is supposed to appear at the same position on each card. Then a mechanical device can assist counting by putting each card from the sorted stack in front of eyes of a counting person for about a second. A light marker may highlight the position for the name of X. In this way the counting person does not have to concentrate on the flow of cards, he/she should just follow continuous appearance of the name of X at the highlighted position. If a wrong name occurs in the highlighted position, the assisting machine should be stopped and the wrong card should be manually examined. These stops are supposed to occur rarely.

Probably, the described checking process can be speeded up so that each card is shown just for a fraction of a second, if other possible names in the place of X are visibly different. (If there are candidates with similar names, their names may artificially be printed with different fonts or at perceivably different position, etc.) The light marker may be momentarily and synchronically switched off and on while the card under view is being replaced by the next one, so that the highlighted name of X would appear continuously as long as there are no erroneous cards. If there is an error, the checking person should notice the momentarily change of the printed record. Then the last few displayed cards should be checked. Officials should test human alertness and adjust the speed of the assisting machine beforehand. [There may be several human checkers or observers, some with capability to stop the machine if an error supposedly occurred.]

Once all sorted votes for the same nomination are checked and sorting errors are rectified, stacks with the same votes should be counted. Counting persons or devices do not need to know what choice is recorded on the cards of a stack offered to them. One way to count the checked cards handily is to load them into container frames of fixed capacity (say, 100). Then counting persons have just to check whether those container frames are filled properly, and to count the container frames. Partially filled container frames should be taken into consideration. If container frames can be firmly stacked in a rectangular fashion, they can be counted by using mathematical multiplication.

It may be acceptable to trust the count of cards with the same vote to a simple mechanical device, reminiscent to a money counting machine. In this way, the only human involvement in the whole process is checking sorted cards. [In case of doubts, counting devices of different type or brand can be used.]

As we see, the three-step counting process can be accomplished with little direct human involvement, yet with full observance and control from them. Involved assisting machines may have simple and unambiguous construction. Laborious human handling or keeping track of lengthy counts can be avoided. Direct human touch is necessary only for dealing with wrongly sorted cards.

If sorting, checking or counting machines are out of order, cards can be manipulated and counted manually, with well pre-specified instructions. Manual count of magnetic cards is already more convenient than manual count of paper ballots.

If the electronic urn is not working, the back up is a ballot box. It may be more compact than a usual wooden box for paper ballots. Manufacturers may provide a foldaway box.

In voting booths, if card readers/printers work well but touching screens fail, the back-up can be regular computers with Windows (or any other OS). The software, including drivers for card readers/printers, can be provided my manufacturers. [As mentioned above, software security at this stage is not critical.]

The most critical scenario is when card readers/printers in voting booths cannot work. Then stickers can be employed. [We still assume that cards record votes for one nomination.] In case of failure of card readers/printers, stickers with a short lists of candidates (and the write-in box) should be prepared fast. The stickers can be printed fast in a central facility with good printers. An A4/Letter page may contain at least 9 stickers to be put on voting cards.

In the troubled precinct, each voter would receive the same voting card, but one side (the one where vote is normally supposed to be automatically printed) covered by a sticker. The sticker should contain a short list of 5-6 names of (presumably) most important candidates. All voters should be provided with alternative voting instructions. In particular, they should be instructed to check first a full list of candidates put in a voting booth. Then they should check whether the name of the chosen candidate is printed on the card sticker, and then mark or write-in that name. This should address satisfactorily the legal issue of discrimination between presumably major and "other" candidates. [Under normal circumstances, all names are equally accessible on a voting screen.] Write-in instances can be expected to be rare.

After the voter made his choice, he should submit the card to a voting box or an electronic urn. Electronic records are not available, so sorting has to be done manually. But sort checking and counting can be automatized. [Checking machines may need to be slower, if only dots have to be followed.]

It may be reasonably assumed that equipment emergencies would not occur in whole districts. (For example, a global power failure would put the election in jeopardy just because basic facilities at polling stations would fail, or voters would have difficulty reaching them.) Then it is not necessary to prepare back-up packages (sticker paper sheets, voting boxes, etc.) for each precinct. A central office or facility may contain a reasonable number of back-up packages.

5. Variations

Magnetic cards can be easily replaced by smart chip cards or similar cards. The essential feature is that cards would be easy to manipulate electronically and manually. The same function can be achieved, for example, by hard paper cards, if reading magnetic strip is replaced by perforation, optical scan or printed bar code.

To reduce costs, it is desirable to be able to use the same cards multiple times. I heard that special ink is used for transportation magnetic cards mentioned in Section 1, so that printed records can be erased with special equipment. (Of course, for voting purposes, printed records should not be erasable by other means.) Other possibility is to use stickers as primary means of automatically printing votes. [Again, the stickers should be removable only with special equipment. Primary and back-up stickers would be printed with different information.]

In the description above, we assumed that vote cards record only one choice for an elective nomination. The cards can record votes for several nominations (or political issues) at once; then several runs of the counting procedure are required if one wishes to apply the full counting scheme for each nomination and issue.

With cards recording multiple votes, the critical back-up scenario can be organized as follows. As a voter-friendly principle, each back-up sticker should list choices only for one nomination. The voting cards may be covered with stickers for two most significant nominations, one for each side. If voting cards should normally record more votes, back-up cards can be used. The back-up cards can be any disused cards of standard format, collected before the election. They would be covered by stickers with voting information for less significant nominations. [A similar possibility is to make and print back-up cards from blank hard paper at a central facility.]

With any back-up scenario, stickers and additional cards should be well monitored.

The voting scheme can be used in priority voting, where voters arrange the choices for the same nomination or issue in preference order. Completely controlled count might be too cumbersome, but a good mix of machine tallies and accordingly chosen "human" recounts could be acceptable.

If consistency of magnetic and printed records is widely trusted, magnetic records can be read and counted as submitted to the voting urn, and the tallies can be accessed and announced as preliminary immediately after the polls are closed. Nevertheless, to keep high level of confidence in the voting system, it is obligatory to perform the count of printed records.

The voting urn may sort cards immediately. Checking and counting machines may recount magnetic records, for the purpose of tracking possible errors.

There may be several ingenious ways to count sorted and checked cards. If fixed capacity of container frames cannot be guaranteed, they should be labeled with the deviation from a fixed medium capacity. Then containers can be counted as with with capacity, and then the total deviation must be added.

Instead of putting cards into container frames of fixed capacity, cards may be automatically grouped into heaps or fans of fixed size. The size of each heap can be checked by a weighting mechanism (say, along a conveyer); fans of fixed form can be checked visually. If a conveyer is used, a sensor along it may count passing container frames or card heaps, and the count may be displayed nearby; then the counting person has just to follow whether the tally is incremented by 1 with each passing item instead of concentrating on the tally number.

Independent or interested observers may estimate the number of votes in any nomination by timing the checking or counting procedures with a chronometer. Interruptions should be taken into consideration. Size or height or card stacks can be quickly compared visually.

6. Conclusions

The proposed voting method is fast and reliable, perhaps more reliable than any other known method. Just as with paper ballot voting, votes have physical form, and voters are ensured that their intended votes are counted. The advantage is that cards are easier to manipulate, and direct human touch can be greatly minimized. There are many ways to assist the human count by rather simple machines. Each step and the whole procedure can be carefully specified, to avoid ambiguous handling and abuse of authority.  

The proposed method avoids highly criticized drawbacks of purely electronic voting, such as the uncertainty whether voter's choice is properly recorded and manipulated. There is no need for checking potentially ambiguous software code, to distrust manufacturers and vendors of voting machines, or to suspect hacking. Correctness of electronic manipulations (by the voting machine in the booth and sorting machinery) is not essential; it is desirable only to ensure smooth counting process. The method does not require awkward paper trails. Accurate results can be highly expected.

The sorting, checking and counting machines may involve marginal amount of electronics, and they can have simple and unambiguous mechanics. The voting method offers favorite conditions for multisided observation. Partisan manipulation of votes should be very difficult. There should be big political interest to implement this method. This may offer big commercial opportunities for manufacturers of voting machines and magnetic cards (and machines dealing with cards).

A whole package of equipment may consists of the following:

• electronic voting machines in booths, with card readers/printers;
  
• the electronic urn;
  
• sorting machines;
  
• machines assisting the checking procedure;
  
• mechanical counting machines (may be optional);
  
• card containers or container frames.
  

The total procedure and equipment collection may look complicated. But each item has a clear function, conceivably simple construction, and should be cheap. What is most important, the system is more reliable, more accurate  and more tamper-proof than even the usual paper ballot voting. This is because direct human manipulation can be minimized, there are ample possibilities for close observation, counting routines can be very well specified. This should make partisan fraud very difficult. The method is also voter-friendly. Solid back-up scenarios make it failure-proof.