SAV Version 1.0

SAV Version 1.0

Copyright © 2009 John L. Ries. Permission to copy, distribute, and modify this program and accompanying documentation is granted under the terms of the GNU General Public License, version 2, as published by the Free Software Foundation, or any later version at the recipient's option.

Contents

  1. What is it?
  2. Where do I get it?
  3. How do I install it?
    1. System Requirements
    2. Building from Source Code
    3. Installing the SAV Executable
    4. Why is it written in Pascal?
  4. How do I use it?
    1. Quickstart
    2. The Control Deck
    3. The Vote File
    4. Results
    5. Technical Support
  5. How does it work?
    1. Initial Tabulation
    2. First Allocation
    3. Second Allocation
    4. Final Allocation
    5. Awarding of Seats
  6. Why would I want to use it anyway?
    1. It is truly proportional.
      1. How is it proportional?
    2. It is Officially Non-Partisan.
    3. It Discourages Polarization.
    4. It Makes Politicians Individually Accountable To the Voters.
    5. It is Resistant to Gerrymandering.
    6. It is Simple to Administer.
    7. Won't This Undermine the Two Party System?
    8. What would this do to my party?
    9. Why are you in favor of proportional representation?
    10. What is wrong with political parties?
  7. Why are you giving it away?
  8. What remains to be done?
    1. Testing.
    2. Weighted Voting.
    3. Write-in Votes.
    4. "None of the Above".
    5. Resolution of Tie Votes.
    6. Friendlier User Interface (GUI?).
    7. More Efficient Assignment.
    8. How to Fill Vacancies.
    9. Other Ideas Welcome.
  9. Sample Data and Scripts
  10. Glossary

What is it?

SAV stands for Single Allocated Vote, a form of proportional representation developed by John L. Ries, beginning around 1995. It is a variant of Single Transferable Vote (STV) that does not involve ranking of candidates by voters. Instead, each voter selects as many candidates as he likes (regardless of the number of seats to be filled) without ranking them. The number of persons voting for each candidate will be used to determine the ranking. Thus, each person's vote will be credited to the most popular candidate voted for, who actually needs that vote to get elected. The number of votes required for election is the Droop Quota, which is equal to the number of persons voting in the election divided by one more than the number of seats to be filled (rounded down to the nearest unit), plus one.

The program SAV is an implementation of this algorithm.

Where do I get it?

SAV is available on Github.

How do I install it?

System Requirements

The program is written in Pascal and thus far has only been compiled with the GNU Pascal Compiler (GPC) under Linux. A makefile is provided which should work on any UNIX or Linux system that has GPC installed. I may make binaries available for selected platforms, should I feel so inclined. Since SAV does compile under GPC, it should be easily portable to any platform supported by GCC (just about everything current, except for IBM mainframes).

I recently attempted a compile under Free Pascal (FPC) which failed miserably. On researching the issue, I discovered that my code wasn't nearly as Borland compatible as I thought. I use Borland style units, but I also use some Extended Pascal constructs not supported by the Borland dialect. Because the Borland dialect is the most widely used, it is my intent at this point to work for full Borland compatibility (as much as I believe in open standards), but I reserve the right to opt for Extended Pascal compliance instead. Either way, my goal is to allow easy compilation on as many platforms as possible, and in any event, SAV will remain fully GPC-compatible for the forseeable future.

Building From Source Code

  1. Edit the makefile to taste.
  2. Type "make"
  3. If it works, you should get an executable named "sav" (or sav.exe on DOS/Windows). Try running the program without arguments. You should get a useful help message.

Installing the SAV executable

This is very easy: just copy the executable into a directory in the path like /usr/local/bin. You will probably want to set the permissions so that everyone can execute it.

Why is it written in Pascal?

Because it makes for more readible code, and is less prone to obscure bugs than is C/C++. While it means that quite a few people would have to install a Pascal compiler in order to build from source code, I anticipate that in the future, GPC will be included in the GCC distribution and will thus be more readily available. If not, I can always rewrite SAV in Ada :-).

How do I use it?

Quickstart

The help message, which is displayed if SAV runs without arguments, is as follows:

SAV 1.0
Tabulates and prints election results according to
Single Allocatable Vote (SAV).

Usage:
sav <control deck> <vote file>

The control deck will be a text file. The first line will contain the number of seats to be filled, followed by the runoff field, which will be 1 if a runoff is permitted, and 0 otherwise. The remaining lines will contain the names of the candidates, one per line in the order of the indices used in the vote file.

The vote file will be a text file, with one vote per line. Each vote record will consist of the indices of the candidates voted for, delimited by spaces.

The Control Deck

The control deck contains the specifications for the election. As stated above, the first line contains the number of seats to be filled and a runoff indicator. The remaining lines contain the candidates running (one line per candidate). An example, with fifteen candidates competing for five seats, follows:

5 1
Aaron Aargau (Air)
Betty Beltran (Air)
Charles Carney (Air)
Daniel Davis (Air)
Earl Effingham (Air)
Fred Feldt (Fire)
Gregory George (Fire)
Henry Higgins (Fire)
Issac Ireland (Fire)
John Jones (Fire)
Kevin Klein (Water)
Lawrence Land (Water)
Mary Mullen (Water)
Norman Nash (Water)
Orville Offenbach (Water)

The record for each candidate can include party affiliation and whatever other data are required, but all are treated as part of the name. SAV does not use candidates' party affiliations to determine allocation of seats.

The "1" after the "5" in the first row indicates that a runoff will be required if fewer candidates receive the minimum number of votes required for election than there are seats to be filled. In this case, twice as many candidates will qualify for the runoff (if there are that many left) as there are seats remaining to be filled. Not surprisingly, the top vote getters not elected outright will be qualify for the runoff.

Should the runoff field be set to "0", any seats remaining to be filled will be awarded to the top remaining vote-getters. It is anticipated that the more seats there are to be filled and the less fragmented the electorate, the less necessary runoffs would be.

The Vote File

The vote file is a text file, will one vote record per line. Each vote record will consist of the indeces of the candidates voted for, delimited by spaces. Each candidate's index is determined by the position of his name in the control deck, thus, if the sample control deck were used, Aaron Aargau is candidate number 1, Betty Beltran is candidate number 2, etc. The indices can appear in any order on the line, as they are unranked.

First few lines of an example vote file:

4 2 1 5 14 11 9 6
11 10 14 4 3 2
1 5 2
10 4 3

In this example, the first voter selected Daniel Davis (4), Betty Beltran (2), Aaron Aargau (1), Earl Effingham (5), Norman Nash (14), Kevin Klein (11), Isaac Ireland (9) and Fred Feldt (6). The next voter selected Klein (11), John Jones (10), Nash (14), Davis (4), Charles Carney (3), and Beltran (2). Both of the first two voters selected more candidates than there are seats to be filled; the next two selected fewer. In any case, each voter has only a single vote which will be credited to only one of the candidates for whom he voted; the fewer candidates voted for, the more limited the choice. Blank votes are ignored. Voting for all candidates has the same practical effect as voting for none (except that a vote is ultimately credited to a candidate).

Results

To run with ctldeck.txt as the control deck, and votes1.txt as the vote file, SAV would be invoked from the command prompt as follows:

$ sav ctldeck.txt votes1.txt

SAV automatically assigns and tabulates the votes, producing a table like the following:

Number of seats to fill: 5
Number of votes cast: 8963
Election Quota: 1494

                     Candidate    Raw vote  Allocation    Status
           Norman Nash (Water)        3056        1494         *
             John Jones (Fire)        3034        1494         *
          Issac Ireland (Fire)        3031        1494         *
          Henry Higgins (Fire)        2998        1369         +
          Charles Carney (Air)        2993         854         +
         Gregory George (Fire)        2974         575         +
           Mary Mullen (Water)        2963         407         +
     Orville Offenbach (Water)        2958         306
              Fred Feldt (Fire)       2937         263
            Daniel Davis (Air)        2935         175
          Earl Effingham (Air)        2934         134
         Lawrence Land (Water)        2932         113
           Betty Beltran (Air)        2927         105
           Kevin Klein (Water)        2902          96
            Aaron Aargau (Air)        2879          84
% votes cast for winning candidates: 67.77

*Elected
+Qualified for runoff

The raw vote is the number of persons voting for the given candidate. The allocation is the number of votes credited to that candidate. The sum of the allocations will always be equal to the number of persons voting.

In this example, three candidates: Nash, Jones, and Ireland, are elected outright. The next four candidates, Higgins, Carney, George, and Mullen qualify for a runoff to fill the remaining two seats. When conducting the runoff, votes are scored in the same way as in the general election, except that the runoff flag is set to 0. The new control deck would be as follows:

2 0
Henry Higgins (Fire)
Charles Carney (Air)
Gregory George (Fire)
Mary Mullen (Water)

And the runoff results might look something like the following:

Number of seats to fill: 2
Number of votes cast: 9190
Election Quota: 3064

                     Candidate    Raw vote  Allocation    Status
          Charles Carney (Air)        4934        3342         *
         Gregory George (Fire)        4934        3114         *
          Henry Higgins (Fire)        4925        1870
           Mary Mullen (Water)        4871         864
% votes cast for winning candidates: 81.03

*Elected

In the first example, if the runoff flag had been set to 0, the top five finishers would have been elected outright. You will notice that the fourth place finisher in the initial balloting (Higgins) ends up losing the runoff to both Carney, and George, who finished fourth and fifth (respectively) in the initial balloting, but first and second in the runoff.

Technical Support

For now, I am willing to answer queries by e-mail (See my Github profile). Should the volume become too great, I may consider starting a mailing list or forum. Not surprisingly, paid consulting gigs and my regular job both get priority over free support. Details are often important; clarity always is. The easier you make it for me to help you, the prompter the help is likely to be.

How does it work?

Initial Tabulation

Since the raw vote totals are used to determine the ranking of candidates, the first step is to count the number of persons voting for each candidate and the number of persons voting in the election. These are used to calculate the Droop quota, which is the number of votes required for election:

Quota:=(NVotes div (NSeats+1))+1;

which is the minimum number of votes necessary to insure that it is mathematically impossible for more candidates to be elected than there are seats to be filled. In the event of an election for a single seat, the quota will be an absolute majority (50%+1) of the persons voting.

First Allocation

The next step is to allocate each of the votes cast to a single candidate. This will be the candidates voted for with the largest raw vote, that has not yet reached quota, and thus has already been elected. If all candidates voted for have reached quota, the vote will be allocated to whichever one has the largest raw vote.

Ties are currently broken in favor of the candidate that appears first in the control deck. It is anticipated that in the future, ties will be broken by a random draw, but at present, the order of candidates in the control deck can be randomized, if desired. In any event, ties are highly unlikely, unless the number of persons voting is small.

If, after this initial allocation, as many candidates reach quota as there are seats to be filled, they are declared elected and no more processing need take place.

Second Allocation

If not all seats are filled, the next step is to transfer votes from those candidates who have already achieved quota to those that have not. This is done by sorting the candidates in order of allocation and then transferring votes (if possible) from the candidate with the highest allocation who has already reached quota, to the candidate voted for with the highest raw vote that has not yet reached quota (if any). The candidates are re-sorted after each transfer. The process stops when either all seats are filled, or there are no more votes that can be transfered.

Final Allocation

If, after the first two allocations, there are still seats remaining to be filled, the next step is to transfer as many votes as possible from the candidates with the smallest allocations to the candidates with larger allocations that have not yet reached quota. This is done by "eliminating" the candidate still under consideration with the smallest allocation and then transferring as many votes as possible from that candidate to the candidate voted for with the highest raw vote that has not yet either reached quota, or been eliminated. If no more votes can be transferred from that candidate, the candidate with the next smallest allocation is eliminated and the process continues. The process ends when either all seats have been filled, or all candidates have either reached quota or been eliminated.

Awarding of Seats

Seats are awarded to the candidates that have reached quota. If, after processing is complete, there are still fewer candidates elected than there are seats to be filled, either the remaining candidates with the largest allocations can simply be declared elected without a runoff, or a runoff can be held between the remaining candidates (up to twice as many as there are seats remaining to be filled) with the largest allocations. Ties in either event are, at this point, resolved first in favor of the candidates with the largest numbers of raw votes, and then in favor the candidates first listed in the control deck. A random draw may be implemented instead in later versions of the program.

For informational purposes, the percentage of those voting who selected one or more winning candidates is calculated and displayed. This will almost always be higher than if the seats were awarded in accordance to plurality, usually much higher, unless the number of seats contested is small (ie. one or two).

Why would I want to use it anyway?

It is truly proportional.

Unlike pseudo-proportional systems such as cumulative voting, or incomplete list systems, SAV guarantees that seats will be awarded to competing parties or factions roughly in proportion to their support among the electorate. Unlike winner-take-all systems (including plurality systems, such as those used in the US, Canada and the UK), SAV makes it nearly impossible for any faction to win a majority in an elective body without majority support (barring unequally apportioned constituencies). Plurality voting allows a party or faction to win a majority with as little as one third of the vote, if the opposition is severely fragmented.

How is it proportional?

Caveat: It is not, unless the number of seats per constituency (district) is sufficiently high.

SAV, like Single Transferrable Vote (STV), seeks to evenly divide the electorate among as many candidates as there are seats to be filled. Since each voter has a single vote, the votes credited to each candidate represent (in theory) a different segment of the electorate.

It is Officially Non-Partisan.

SAV completely ignores party affiliations of candidates. This allows voters to divide their votes among candidates of competing parties as they see fit. It also allows independent candidates to compete on the same basis as partisan ones. SAV can even be used in officially non-partisan elections, and will award seats among whatever factions (to include social classes and religious and ethnic groupings) divide the voting population, even if they cut across formal party groupings, and even if formal party groupings do not exist. It is even appropriate for use by clubs, corporations, and other private organizations where party affiliation is, or should be irrelevant.

It Discourages Polarization.

By awarding votes to the candidates with the broadest support in each faction, SAV rewards efforts to appeal to voters across party and factional lines. Conversely, candidates with narrow bases of support are penalized. This is in contrast to STV (Single Transferrable Vote), which tends to favor the candidates within each faction with the most enthusiastic followings, and party list systems, which tend to reward party loyalty and divide the electorate along party lines, discouraging cooperation between politicians and voters of different factions on matters where they agree, except as necessary to secure majority support for a given proposal, which is often difficult.

Furthermore, plurality at-large systems, which are often used in local government, often increase polarization by making it possible to shut out the opposition, thus putting a premium on cut-throat tactics. SAV, by making it more difficult to shut out minority factions, will tend to reduce the motivation for cut-throat politics, thus allowing politicians and voters with differing opinions to more peacefully coexist.

It Makes Politicians Individually Accountable To the Voters.

Since voting is for individual candidates, instead of for party slates, candidates are directly accountable to the voters, instead of being primarily accountable to their parties. The biggest problem with slate voting is that it makes elected officials more accountable to their parties than they are to the voters they theoretically represent. The problem is compounded by closed list systems that allow parties to order the candidates by preference, thus rewarding party loyalty and seniority, and making the individuals elected almost completely unaccountable to the general public.

It is Resistant to Gerrymandering.

In common with other proportional systems, if the number of seats per constituency is sufficiently large (perhaps 5-7), seats will be awarded to parties or factions in proportion to their support among the electorate, frustrating efforts to rig elections in favor of one faction or another through the creative drawing of election districts (often a danger with majority or plurality systems).

It is Simple to Administer.

Balloting takes place in the same way as in a plurality election, except that no limit on the number of candidates voted for need be imposed. Elections conducted on the basis of single transferable voting (STV) require voters to rank the candidates, and it often takes days before results are known. Assuming that ballots are in a form that is easy to scan electronically, votes could be rapidly read for processing by SAV, and results published the same day.

Won't This Undermine the Two Party System?

Maybe. Plurality voting systems strongly discriminate against independent and minor party candidates. Voters are discouraged from supporting such candidates on the theory that they "can't win", and that any votes for them are "wasted". On the other hand, I doubt that a two party system that survives only through intimidation is really worth having. Furthermore, there is nothing in SAV that prevents the electorate from voting overwhelmingly for major party candidates, should that be their true inclination.

In addition, plurality voting systems are not always enough to prevent the splintering of the electorate among multiple parties. Although one party almost always wins a majority of seats in the British House of Commons (and thus, the privilege of forming the government), that party almost never has majority support among the electorate. Indeed, while the Conservative and Labour Parties together normally win the vast majority of the seats, and even a comfortable majority of the votes, the Liberal Party (now the Liberal Democratic Party) has always attracted considerable support and at least a few seats (now a substantial number). In addition, regional parties now command the support of a growing number of voters and win increasing numbers of seats. All this is true in spite of the fact that elections are conducted by plurality voting in single member constituencies. The same trend is occuring in Canada, where four parties now routinely win seats in the House of Commons, despite the fact that members are elected by plurality vote from single member constituencies. As of this writing (April, 2009), the last three federal Parliamentary elections in Canada have resulted in minority governments.

Even in the USA, candidates are sometimes elected over the opposition of most voters because the opposition is divided. Elected officials chosen on this basis often enjoy less legitimacy in the eyes of voters than those with majority support, and are therefore often less effective. Furthermore, I suspect that efforts to win elections by dividing the opposition, instead of by winning majority support, tend to undermine public confidence in elections, thus alienating the people, without whose support representative government is impossible. Again, even in the USA, it is becoming more common for no candidate to win a popular majority in the presidential election, due largely to increased dissatisfaction with the major parties. In three of the last five US presidential elections (1992-2008), the winning candidate was elected with less than a majority of the popular vote.

Another problem that often occurs in single member constituency systems is that many constituencies are "safe" for one party or another. In such cases, the nomination of the dominant party is tantamount to election, discouraging voter participation and reducing accountability. In addition, the once disreputable practice of gerrymandering has gained great favor in the USA in recent years. As of the time of this writing, election districts for the US Congress and state Legislature in California are so drawn as to virtually insure the election of the candidate of a particular party in nearly all of them. This was deliberately done to minimize the possiblity of seats shifting from one party to another.

What would this do to my party?

My prediction and hope would be that replacing a majoritarian electoral system with SAV would make it much more difficult for a dominant party to retain that status in the long term (either globally or locally). It would also be my expectation that it would tend to relax party discipline by making candidates less dependent on their parties to win election. As noted above, any system of proportional representation would also make it easier for candidates of smaller parties to be elected, which would make it more difficult to maintain a two party duopoly such as the Democratic/Republican rivalry that has dominated US politics since the 1850s. It would also be my expectation that it would make it more difficult to exclude candidates representing ethnic or religious minorities from serious consideration, even in the face of major party collusion.

That said, how well a particular party would do under SAV would be largely dependent on its ability to win and maintain the confidence of voters. To the extent that people trust a particular party to promote the public interest, that party will have an easier time winning election of its candidates and adoption of its policies, than otherwise. To the extent that parties rely on being the "only game in town", or the only realistic alternative to the "Bad Guys", they will tend to suffer under a system that affords voters a wider variety of viable candidates. Smaller parties will tend to do better under proportional representation than under a majoritarian system, but the degree will depend on their ability to win the confidence and allegiance of mainstream voters and the ability of major parties to "hold their own".

Why are you in favor of proportional representation?

I am not, entirely. I think there is much to be said for majority voting, particularly the fact that it forces candidates to appeal to the majority of the electorate, rather than to a narrow fraction. Majority voting is also a lot simpler and more intuitive than are most systems of proportional representation, and makes it easier to hold individual politicians accountable than does any system of proportional representation, SAV included. Combined with single member constituencies, it greatly reduces the number of elected officials of whom voters need to keep track, allowing such voters to make more informed judgments regarding individual politicians.

That said, I think a nonpartisan system of proportional representation, such as SAV, offers a number of advantages, especially when the electorate is polarized on party, class, regional, or ethnic lines, or in situations where one party is dominant. I think SAV would be an enormous improvement over the plurality at-large system typically employed in municipal elections, or the plurality multimember consituency system employed in elections to some state legislative bodies (such as the New Hampshire House of Representatives). As noted before, proportional systems are resistant to gerrymandering, and make it easier to elect representatives on a the basis of natural regional groupings, rather than from arbitrary constituencies for which numerical equality is the primary, and sometimes the only basis.

Possibly, it would make sense to elect one house of a bicameral legislature from single member constuencies by majority vote (instead of plurality), and the other on a regional basis using SAV, or to elect a unicameral assembly partly one way, and partly the other. In any event, SAV should be seen as one of a variety of methods that might or might not be appropriate depending on the circumstances, rather than as the optimal system in all cases. It is one tool in the toolbox, not a magic bullet.

What is wrong with political parties?

Having read this far, you might get the impression that I am less than sympathetic to partisan politics and you would be correct. I wrote an essay several years ago in which I discussed at length the reasons for this aversion. Suffice it to say here that I view political parties as unnecessary intermediaries between the people and their elected leaders, that they enforce false dichotomies by packaging together sets of political beliefs and policies that may or may not be related to each other, that they encourage emnity between people who might otherwise be encouraged to cooperate on matters of common concern, that they too often put the effort to win elections ahead of the public interest (often engaging in corrupt practices or tolerating corrupt officials), and far too often, have (deliberately or not) restricted freedom of thought and conscience by making conformity to a predetermined orthodoxy the price for meaningful participation in the political system (if not for retaining one's employment, staying out of prison, or maybe even staying alive). Political parties may be inevitable (or not) in a system of government based on free elections, but at their best, they are advocacy organizations formed by groups of like minded citizens for the purpose of promoting particular policies and candidates for public office. This role does not require that they control the nomination of candidates for office, that parliamentary bodies be organized along party lines, or that executives make party affiliation or loyalty leading criteria for appointing or dismissing their subordinates. Let parties make such recommendations to voters and elected officials as they think best, and then allow them to follow their own consciences without fear of recrimination. Parties do not need and should not have a formal role in either the nomination of candidates or in the operation of governmental institutions, which should represent all citizens, instead of just the majority party or coalition.

Why are you giving it away?

While I believe that it is appropriate (though certainly not mandatory) to sell most software on a proprietary basis, software that performs sensitive public functions (like scoring elections) requires a degree of public confidence that can only be achieved by opening the source code to public inspection. The use of open source software to score elections eliminates a major potential source of election fraud, error, and controversy. Such will always be a danger with closed source software, no matter how honest the developer, or the administrator.

In addition, my primary purpose in writing this program is to present the SAV algorithm to the public. Widespread adoption, or at least, discussion of this algorithm is, in this case, sufficient reward for my efforts, though monetary contributions will certainly not be refused.

What remains to be done?

Testing

The "right way" to test this algorithm and to assess its impact on elections where it is used, would be to test it on real election data. The problem is that combinations of votes cast in previous elections are rarely, if ever, saved, particularly here in the USA, where nearly all elections are decided by either plurality or majority vote. I have created some artificial data sets, which attempt to simulate voting with varying degrees of partisanship. Not surprisingly, where party loyalty is weak and ticket splitting endemic, results tend to be similar to those obtained by plurality voting. As partisanship becomes stronger, results look more like those associated with free list proportional representation. That is what I expect.

One line of research I intend to pursue is to determine the extent to which the order in which votes are counted affects the results. I expect the effect to be minor, particularly if the number of votes is large, but if it is significant, then steps would need to be taken to mitigate it.

Results from previous elections in California (and presumably other states) are available aggregated by precinct. Such data could be used to create artificial voting data with by-candidate totals matching those recorded. Such data could be constructed with varying assumptions about the level of partisanship. I have not attempted to use such data, but I hope to do so in the near future.

Should "real" data be available somewhere (maybe from STV jurisdictions like Ireland or Australia), I would love to try SAV on them. Such real data could also include results from elections to the governing board of your favorite club or volunteer organization.

Someday, should I have time to test SAV adequately, I might even publish a paper on my findings. In the meantime, I would love to read any research others might do.

Weighted Voting.

For better or worse, voting in for-profit corporations is conducted on a weighted basis, each voter typically having as many votes as he has shares of stock. I believe that corporations could benefit by affording a broader range of stockholder representation on their boards than is now usually done (interest in such things even crops up in the news occasionally). Since, under the usual winner-take-all system, corporate boards are nominated by a controlling faction, and ratified by the stockholders with little or no opposition, it is easy for such boards to become rubber stamps for management, which they often are. A more broadly representative board of directors could more effectively supervise the management, and could insure that minority concerns are at least discussed. Such could be true, even if one person has a controlling interest. The majority still wins, but at least the minority gets to participate in the discussion.

Write-in Votes.

It should be possible to write in the names of persons for whom one wishes to vote, either because none of the ones listed are acceptable, or others are deemed more suitable. This could be done by allowing a write-in code (maybe the number of official candidates, plus one) in the vote file, to be followed by the name of the person "written in".

"None of the Above".

While desire to vote for "none of the above" is symptomatic of insufficiently competitive elections, and should be lessened by the greater competition afforded by systems such as SAV, it may still be a good idea to offer the option as a way to "keep the big boys honest". Under SAV, a vote for "none of the above" (probably represented by a zero), combined with no votes for actual candidates, would increase the number of votes cast, without increasing the number of votes recorded for actual candidates. This would have the effect of raising the quota and thus making it harder to win election outright. If "none of the above" were to receive enough votes to reach quota (assuming a single round election), a proportionate number of seats would be declared vacant, and a new election would have to be called to fill them, from which all candidates running in the current election would be disqualified. In a two round election, greater numbers of "none of the above" votes would make it more difficult for the leading candidates to reach quota, but the candidates qualifying for the runoff would have an additional opportunity to win over dissaffected voters before seats are declared vacant, or existing candidates are disqualified. Of course, a "none of the above" vote should be ignored, if combined with votes for existing candidates

In general, it is much more constructive for voters who fail to find at least one suitable candidate on the ballot, or to write in the names of one or more persons they do find suitable, than it is for them to simply reject all candidates outright. Thus, the penalty for large numbers of people voting for "none of the above" should be a proportionate number of seats being left vacant, pending a new election. That way, a vote for "none of the above" could be properly interpreted as preferring one or more vacancies to the election of any of the existing candidates.

Resolution of Tie Votes.

I expect ties to be rare, but not unheard of. At present, ties are resolved by ranking candidates in the order they appear in the control deck. A random ordering could be used instead, or a random draw could be made at the time the tie occurs. There may be other possibilities.

Friendlier User Interface.

The existing command line interface is portable, relatively easy to program, and easily incorporated into scripts, but interactive, graphical user interfaces (GUIs) have become the standard. Indeed, most computer users now have little or no experience with command lines or scripting, and are comfortable only with a GUI. A GUI front end with the ability to specify candidates and parameters, and possibly even to enter the votes, might well be helpful for such users.

More Efficient Assignment.

It may be possible to reduce or eliminate the need for re-assignment of votes if they are sorted properly. One idea is, after tabulating the raw votes, sorting the votes in reverse order of the popularity of the candidates voted for. Thus, votes cast only for the least popular candidate would go first, followed by votes only for the second least popular candidate, followed by votes for both of the two least popular candidates, etc. Votes cast for all candidates (assignable to any of them) would go last. Another possibility would be to rank in ascending order of the number of candidates selected (singleton votes first). In either case, the least flexible votes are allocated first. There may be other sorting orders that would work better.

How to Fill Vacancies.

While it is certainly possible to fill vacancies by simply holding a special election, it would, in all likelihood, throw proportionality out the window. Therefore, it seems a good idea to devise a mechanism for filling vacancies more in keeping with the spirit of SAV. One possibility would be to recount the ballots from the previous election with the person whose removal caused the vacancy, plus any other no-longer-eligible candidates disqualified. Votes for candidates elected and currently serving would be counted and credited as they were before. The purpose of this would be to simulate as closely as possible the previous election as it would have been if the departed members and other now-inelligble persons had not been candidates. If one of the previously unsuccessful candidates reaches quota, he/she would be declared elected, otherwise, a special election could be held between the two candidates not previously elected that have the largest allocations. In the latter case, it may be necessary to require a minimum number of raw votes to be eligible (maybe quota or a large fraction thereof). If the vacant seat cannot be filled in either of those fashions, an "open" special election could be held. To accomodate such recounts, it would be useful to be able to instruct SAV to ignore certain candidates, so as to not have to modify the vote file.

A third possibility would be to declare an alternate for each person elected who would be the candidate not elected with the most votes in common with the one who was (the same person might end up serving as an alternate to multiple members). If for whatever reason, the person originally elected did not finish his term, the alternate would serve for the remainder. This would have the advantage of not requiring a special election and being easier to compute than a full rerun of the vote counting.

Other Ideas Welcome.

I certainly do not have a monopoly on good ideas, and am willing to entertain any suggestions for improvement.

Sample Data and Scripts

I have created some data for testing purposes that I include in the distribution.

Control Decks

ctldeck.txt
The control deck used in the first example in this documentation and with most of the included vote files.
ctldeck2.txt
The control deck used in the runoff example in this documentation.
ctldeck2a.txt
Same as ctldeck.txt, except that runoff is disabled (assuming a single ballot).

Scripts, vote files, and output

I created ten vote files, each representing a different voting scenario. For each file (votes*.txt), there is a Perl script that created it (votes*.pl), an output file showing the results (votes*.out), and in most cases, a short file showing the proportion of votes for our three hypothetical parties (votes*.split). Assuming you have Perl installed, you can recreate a vote file by rerunning the script. For example:

$ perl votes3.pl >votes3.split

It is recommended that votes1.txt and votes2.txt not be regenerated, as they are the examples used in this document.

votes1.txt
The basic scenario shown in this document. Five candidates from each of three parties (fifteen in all) competing for five seats. 10,000 voters, each voting for between one and ten candidates, determined randomly. The candidates voted for are determined at random, each being equally likely, regardless of party.
votes2.txt
Runoff election following votes1.txt (shown in this document). The control deck used is ctldeck2.txt. The script votes2.pl assumes that each person who voted for one or more finalists on the first ballot will also vote for them on the second ballot. Otherwise, candidate selection is purely random, with each voter selecting between one and three candidates.
votes2a.txt
Alternative runoff setup using ctldeck2a.txt. Here, the candidate indices are the same as they were on the initial ballot.
votes2b.txt
Alternative runoff setup using ctldeck2a.txt. Here, the only votes cast are those cast for runoff finalists on the first ballot.
votes3.txt
Assume that the probability of a voter aligned with a party supporting any given nominee of that party is 95% and that 10% of voters are independent and will vote without regard for party affiliation. Party alignment for each voter is determined by random draw, and then candidates voted for by each person are determined. This, and all subsequent voting scenarios use ctldeck.txt.
votes4.txt
Drawn in the same fashion as votes3.txt, except for the additional assumption that the probability of a given party aligned voter voting "straight ticket" is 30%.
votes5.txt
Drawn in the same fashion as votes4.txt, except that the probability of a "straight ticket" vote is increased from 30% to 50%.
votes6.txt
Drawn in the same fashion as votes5.txt, except that base party alignment probabilities are determined by random draw (no longer equal).
votes7.txt
Drawn in the same fashion as votes6.txt, except that the party loyalty probability is reduced from 95% to 50%.
votes8.txt
Drawn in the same fashion as votes7.txt, except that the "straight ticket" probability is reduced from 50% to 25%.

Feel free to modify the scripts for the purpose of creating your own scenarios, or write your own.

Glossary

Proportional Representation

One of a class of election systems designed to choose candidates from different parties or other fractions of the electorate in proportion to the number of votes their candidates received. Western European and Latin American parliamentary elections almost always employ some form of proportional representation in multimember constituencies (or at large), sometimes mixed with plurality or majority voting in single member constituencies. Such systems are more representative of public opinion overall than are majoritarian ones, but usually at the cost of political fragmentation, increased partisanship, and reduced individual accountability of those elected (which SAV is designed to minimize). In most representative bodies chosen by proportional representation, it is rare for any one party to hold a majority of the seats, requiring parties to form coalitions in order to govern effectively, or even to pass controversial legislation (due to increased partisanship).

Plurality Voting

The practice of declaring elected the candidate(s) with the largest number of votes, even if supported by only a minority of voters. This is most common in English speaking countries, but is also frequently used elsewhere. It effectively discourages voting for candidates deemed unlikely to win plurality support, thus discouraging political splintering, but also often leading to rule by the largest or best organized minority, rather than by the majority. Plurality voting is most often done in constituencies where only one person is elected at a time, but is also sometimes employed in multimember constituencies, and often at large, such as in US municipal council elections. Plurality voting systems usually require that voters select no more candidates than there are seats to be filled, but this is not strictly necessary, and systems allowing votes for more candidates than there are seats to be filled have been proposed as a way of promoting the election of those candidates in whom the public has the most confidence.

"First past the post" (FFTP) is the practice of electing one candidate (or slate of candidates) per constituency at a time by plurality vote, each voter being permitted to vote for only one candidate (or slate). Under Block Voting, there are multiple seats to be filled, each person beling allowed to vote for as many candidates as there are seats to be filled, and the candidates with the most votes being declared elected. Block voting is typically employed in electing US state and territorial legislators from multimember constituencies, as well as in choosing US municipal councillors at large.

US Presidential Electors are, in nearly all states and in the District of Columbia, elected as slates of candidates pledged to vote for a particular pairing of candidates for President and Vice President. The slate that wins the largest number of votes in a given state is normally awarded all that jurisdiction's Electors. The Electors then formally elect the President and Vice President by majority vote, but the result is almost always a foregone conclusion.

Majority Voting

The practice of requiring a majority of votes for election. If fewer candidates win majority support than there are seats to be filled, a runoff is usually held between the highest remaining vote-getters. This is most common in France and eastern Europe, and in Latin American presidential elections. If a majority is required for election, voters feel more free to support their most preferred candidates (at least in the first round), even if they are deemed unlikely to win, while the sort of political splintering common under proportional representation is still discouraged, and candidates typically run as individuals, instead of as party slates. Unlike plurality voting systems, majority voting systems make it difficult for one party to win a majority of the seats in a representative assembly without majority support among the voting population. However, runoffs are often required, increasing the time and cost of administering elections.

Under Alternative Voting, voters are permitted to rank the candidates in order of preference and if no candidate receives a majority of the first place votes, the votes of the least popular candidates can be transfered to more popular ones in accordance with the wishes of individual voters. Thus, it is a special case of Single Transferrable Vote in which only a single candidate is elected per constituency. The President of the Irish Republic and members of the Australian House of Representatives are elected in this fashion. This procedure is often called Instant Runoff Voting (IRV) in the USA, where it is sometimes advocated as a replacement for plurality voting.

Incomplete voting systems (Limited Voting)

Election systems in which voters are only allowed to select candidates to fill a fraction of the seats to be filled. There are two major variants:

  1. Those where voters select individual candidates. An example is the system for electing at large members of the District of Columbia (Washington, DC) Council: Each voter selects one candidate, but the top two vote getters are elected (two of the four at large councillors are chosen every second year) and no party is permitted to nominate more than one at large candidate per election, making it impossible for any one party to hold more than eleven of the thirteen seats. When voters are only permitted to select one candidate each, but multiple seats are filled, the practice is sometimes referred to as Single Nontransferrable Vote (SNTV). Eleven members of each house of the Puerto Rico Legislative Assembly are elected at large using SNTV.
  2. Those where voters select party slates, the one with the largest number of votes being granted a majority of the seats, and the one with the second largest number of votes usually winning the rest. An example is the system for electing the three fourths of the Mexican Senate chosen by state: Three senators are elected from each state, and the Federal District, but each party is permitted to nominate only two candidates, and voting is for party lists, rather than for individual candidates. The party with the largest number of votes in the given state is awarded two seats, while the one with the second largest number of votes wins the remaining seat. This practice is called an incomplete list system.
Cumulative Voting

An election system in which each voter gets a single vote which he may divide among as many candidates (regardless of party) as desired, and the candidates receiving the most votes are elected. The normal consequence is to insure minority representation in each (multimember) constituency by making it much more difficult for one party to win all of the seats. At one time this was often used in local elections in the USA, and also in elections for the Illinois House of Representatives (each of the 59 legislative districts elected one Senator and three Representatives), but it has fallen out of favor. It is still used in some localities.

Closed List System

A proportional representation system in which each political party or faction nominates a list of candidates in order of preference, and each voter chooses one of them. Seats are then awarded to each party list in proportion to the number of votes it received. There may be a minimum percentage of votes required to receive seats, as is the case for the one quarter of deputies to the German Bundestag that are elected by state, but this is not always the case. Extreme examples are elections to the Israeli Knesset (120 members elected at large), and the South African National Assembly (400 members elected at large). In neither of these two cases is a minimum percentage of votes required for representation, meaning that a party that wins 1/400 of the vote in a South African general election is entitled to a seat in the National Assembly. The closed list system is the single most common form of proportional representation.

Open List System

A proportional representation system in which voters are permitted to select individual candidates within a party list. Seats are awarded to the parties in proportion to the votes each received, and are filled by the candidates from each party that win the largest numbers of votes. There is often a minimum percentage of votes required for representation. This form of proportional representation is common in much of continental Europe and Latin America.

Free List Proportional Representation is a special case of the open list system that allows voters to choose as many individual candidates as there are seats to be filled, the votes being regarded as cast for both the candidates and their parties. This allows "split ticket" voting, which is generally precluded in list systems.

Single Transferrable Vote (STV)

A proportional representation system in which each voter selects one or more candidates in order of preference. Each vote is initally credited to the most preferred candidate of each voter, but can be transferred to candidates lower down on the list, if the more preferred candidates already have enough votes to win election. Typically, the Droop quota (the total number of votes divided by one more than the number of seats to be filled, plus one) is required to insure election. If fewer candidates "make quota" than there are seats to be filled, then the remaining candidates with the highest numbers of votes are declared elected. When STV is used in single member constituencies, it is sometimes referred to as Instant Runoff Voting (IRV) or Alternative Voting (AV). STV in multimember constituencies is the only form of proportional representation in common use in which candidates run solely as individuals, rather than as members of a party slate. It is is rarely used outside of the former British Empire.

STV is widely used in Ireland and Australia, and is traditionally advocated by the Liberal Democratic Party in the UK, where plurality elections are the norm. Dail Eireann (the lower house of the Irish Parliament) is elected by STV from multimember constituences, as is the Australian Senate. 45 of the 60 members of the Irish Senate (Seanad Eireann) are indirectly elected through STV from among panels of candidates nominated to represent different areas of expertise.

SAV is proposed as an alternative non-partisan system of proportional representation and is partly based on STV.

Constituency

A geographical district entitled to choose one or more members of a representative assembly. In the USA, they are usually called "districts", or in certain older cities, "wards".

Comments