About the DemoChoice Web Poll

What's the basic idea?
DemoChoice web polls are designed to produce satisfactory representation for everyone, with majority rule.

If your favorite candidate has too few votes to win, your vote will be transferred to your next favorite, if possible.

If your favorite candidate has more than enough votes, some ballots may be partially transferred so that all winners represent roughly equal numbers of voters.

What is DemoChoice for?
In a democracy worthy of the name, everyone's voice is heard (or represented with their explicit consent), and decisions require at least majority support: more people should support an idea than oppose it.

If you elect your representatives by majority vote, and they make decisions by majority vote, a small group can overrule the will of almost 75% of voters - and up to half of all voters don't even have representatives who will express their protest.

In the United States and Canada, we usually elect people by "most votes wins" instead of majority, so it can be even worse. And worse than that, the people in power can group you with others who will vote against your favorite - they can decide which voters gain representation. No wonder so many people have lost faith and don't bother to vote: this approach miserably fails to meet our goals.

But it can be done! DemoChoice gives you the freedom to express your preferences in detail among many viable choices, and then counts your votes in a way that pursues the democratic goals noted above. It can usually accommodate almost everyone. As a result, voting actually becomes a fun, positive, and rewarding experience!

How does DemoChoice pursue its goals?
DemoChoice attempts to assign everybody to their favorite representative. To make this work, a few adjustments need to be made.

How are the results tallied?
It's easiest to understand this by just watching how the votes move on the results pages, but here are the detailed rules for the count.
  1. Voting
    Voters rank candidates in order of preference, with the understanding that if circumstances prevent any of their votes from counting for a higher choice, they will count for a lower choice.
  2. Counting votes
    Each ballot has a number of votes equal to P (the “precision”), which is 1 for a single winner contest, or otherwise the number of seats multiplied by one more than the number of seats. P can optionally be a whole-number multiple of this. Equivalently, each ballot could be described as having one vote that can be split into fractions with P as the denominator. As described here, a candidate can be a remaining candidate or an eliminated candidate. Remaining candidates include winners and unelected candidates who have not been eliminated. Votes from each ballot are assigned to its highest-ranked remaining candidate, constrained as follows:
    1. Each remaining candidate has a prescribed maximum number of votes per ballot, initially equal to P. For example, if the maximum of one’s first choice is P-1, then the ballot would count P-1 for the first choice, and 1 for the second choice.
    2. If there are leftover votes after applying constraint A, they are reassigned to their lowest-ranked winner if one exists, regardless of the maximum.
    3. Ballots that cannot count votes to any remaining candidate are counted for “none of these”.
    For each winner, a record (histogram) is kept of how many transferable ballots count a given number of votes for each winner. A ballot is transferable if there is a lower-ranked candidate who has not been eliminated. The histogram is kept sorted from highest to lowest number of votes per ballot.
  3. Identification of winners
    1. The winning threshold is calculated as the total number of votes for continuing candidates divided by one more than the number of seats.
    2. Any candidate with more votes than the winning threshold is declared a winner. (Equivalently, the threshold could also be defined as being one vote more, and a candidate exactly meeting the threshold is also declared a winner.)
    3. If there are not more remaining candidates than seats to be filled, all remaining candidates are declared winners.
  4. Maximum reduction
    1. If at least one unfilled seat remains: the histogram of each winner is consulted to determine the lowest maximum that the winner could have while staying above the winning threshold, and that reduction is applied. A proposed maximum can be checked by multiplying the number of ballots in each histogram bin by the difference between its number of votes and the proposed maximum, discarding any negative values, and summing these. The correct maximum can be quickly found by proposing it to be the number of votes per ballot in successive histogram bins starting with the second highest (or zero if the end of the histogram is reached). If this proposed maximum is too low, a proportional (interpolated) and rounded-up value between this and the previous proposed maximum is the correct value.
    2. If only winners remain, and all were declared winners earlier than this round, the “ideal” threshold is calculated as the total number of votes for the winners divided by the number of seats. The histogram of each winner is consulted to determine the lowest maximum that the winner could have while staying above the ideal threshold, and that reduction is applied, following the same procedure as in 3A.
  5. Elimination
    If no maximum reductions are possible in step 3, the unelected remaining candidate with the fewest votes is eliminated. The candidate’s maximum is set to zero. A tie in this result is resolved by attempting to identify which tied candidate had the fewest votes in successively previous rounds that elected or eliminated a candidate, or randomly if this is not successful.
  6. If a candidate was newly elected in step 2, or a maximum reduced in steps 3 or 4, return to step 1. Otherwise, the count is complete.

Sometimes this procedure results in several rounds of maximum reductions before a candidate is elected or eliminated. It can be easier to fathom the results if these are lumped into a single round. More specifically, a new round would be reported only for those with a new winner; those with a new elimination; rounds just after an elimination (because the threshold may have been reduced); when step 3B is first used; or in a final round summarizing all of the winners. Results are typically shown with everything divided by P so that each voter has one vote.

For step 1, a noteworthy choice of the whole-number multiple for P is to specify that the minimum number of votes per ballot must be 55. This way, contests with 2 to 5 seats have 60 votes per ballot and those with 6 to 9 seats range from 56 to 90. This makes the completeness of surplus transfers (as discussed below) less dependent on the number of seats.

Hey! This is too complicated!
The rules behind DemoChoice appear complex, but only because they put nearly all of the electoral controls within reach of the voter. With currently used methods, the outcome of most elections is determined primarily by political consultants who use sophisticated computer algorithms and large databases to manipulate district boundaries and reduce competition. Casting a DemoChoice vote is straightforward, but with current methods, voters must fret over strategy to avoid wasting their vote on a loser or on someone who would win anyway. Please don't give up!

In a multi-winner election, how do you choose which votes stay with a winner?
Transferring votes from winners ensures that all winners represent constituencies of similar size, and that people don't avoid voting for popular candidates, thinking that they will get elected anyway.

Different versions of this counting method do this in different ways. They can be randomly chosen, taken as a fraction from each ballot, or chosen based on the distance between the winner's home precinct and the voters' precincts.

If we minimize the number of times ballots are split into portions that count for different candidates, we establish a clearer connection between voters and legislators. DemoChoice accomplishes this by giving transfer priority to ballots that count larger fractions toward the winner. A maximum fraction of a ballot is determined that transfers a maximum number of votes without putting the winner below the Enough or Ideal threshold, as appropriate. This makes it likely that most ballots will be sliced, but few will be sliced more than once.

What happens if there is a tie?
Ties are not a very significant issue in public elections, because the number of ballots is large and ties are statistically rare. However, in a demonstration poll like this, they can happen frequently, especially among unpopular candidates. Ties are only an issue during eliminations, and usually involve candidates with few votes that do not affect the course of the election. Here, ties are broken by considering who has the fewest votes in successively previous rounds that elect or eliminate a candidate (rounds that only reduce maxima are not considered). If this does not break the tie, a candidate is chosen by the server's random number generator.

Is this the same as Instant Runoff Voting?
Yes, if there is one winner. This method works well for electing mayors, governors, or presidents. IRV usually stops when two candidates remain, but IRSA eliminates the last losing candidate as a clearer measure of depth of support for the winner.

The multi-winner version should be used for boards, councils, and legislatures. This gives more people representation than the usual method of dividing voters into districts and using single-winner elections in each.

How well does it work?
DemoChoice can routinely assign more than 90 percent of voters to representatives they support. This usually means that a decision by a majority of representatives reflects the will of a majority of voters. Winners receive nearly equal shares of votes, so that each vote corresponds to a nearly equal amount of legislative power. Each representative has the unanimous support of his/her voters. Voters have a large number of options because there is no appreciable 'spoiler' or 'vote-splitting' effect to scare away candidates. See for yourself by looking at the results pages on the DemoChoice site!

Where did you get this newfangled idea?
The basic concepts of this method of voting were first proposed in 1821, within a generation of adoption of the US Constitution. Similar methods were proposed independently in the US, Britain, and Denmark, and were used in a few public and private elections in that century. John Stuart Mill, the most well-known scholar on the theory of representative government, tried unsuccessfully to enact it when he served in the House of Commons. Australia, Ireland, Northern Ireland, and Malta have used this method since the early 20th century. Scotland has used it in local elections more recently. The closest relative of IRSA is the method proposed by Hugh Warren in 1983, which combined the concept of maxima with the iterative surplus transfers of Brian Meek's 1969 version. New Zealand adopted Meek's method in 2004 for some local elections.

About two dozen US cities including New York and Cincinnati elected their city councils this way in the first half of the 20th century. It was very effective, but the principle of an equal voice for all was ahead of its time - women had only just been allowed to vote, and this was well before the civil rights movement - so it was repealed in almost all cases. The only remaining case from that era is Cambridge, MA. In 2002, San Francisco adopted instant runoffs to determine a majority winner for mayor and other offices. The nearby cities of Oakland, Berkeley, and San Leandro soon followed, as did Minneapolis (including some multi-winner contests), St. Paul, Santa Fe, and a few other US cities. In 2018, Maine became the first US state to hold statewide instant runoffs for a few offices.

Let's do this in our local, state, and federal governments!
If you are interested in promoting this method of voting, Fairvote can provide more information and help you find like-minded people. Also, browse the DemoChoice library.

How can I print (or save) the results?
To print the bar charts, you may need to change your browser settings to enable printing of background colors. For example, in Microsoft Internet Explorer, choose "internet options" from the "tools" menu, go to the "advanced" tab, and check the "print background colors and images" box. If the dotted threshold line doesn't print, add "&thickdot=on" (without the quotes) to the page's web address.

If the poll has a large number of candidates, the results may be broken into pages. To disable this in order to save or print results, use "&page=0" (no quotes) in the page's web address.

Why didn't the totals change after I voted?
They did - try pressing your browser's 'Refresh' button.

I still don't get it!
We want to make sure that everyone who uses this site leaves with a comfortable understanding of how it works. Please feel free to ask a question. Our library has many links to other explanations and discussions where you can learn more.

What do you do with my email address in a private poll?
Your email address will be used to send a confirmation of your vote. In the rare event that your vote is not properly recorded, you may be contacted. Voter address information is not used for any other purpose.

Send us your feedback!
DemoChoice is an ongoing project, and user feedback is an essential part of it. Everybody has a slightly different experience and it helps to hear what parts you found illuminating and what parts you found confusing or cumbersome. Please share your thoughts!

Acknowledgements

Steve Willett created the first web-based instant runoff poll in 2000, as an interface to ChoicePlus Pro. DemoChoice evolved from this into its own project. Steve and the Center for Voting and Democracy helped provide web space for the first two years. Many others have provided helpful advice and encouragement. Further comments would be appreciated.

DemoChoice Web Polls ©2001 Dave Robinson