PAL (Proportional, Accountable, Local) representation is a system for electing a proportionally-representative legislature. Most voters would vote for a single candidate listed on their local ballot. The vote totals are compared to the threshold for election to a seat, allow for up to one seat worth of leftover votes; for instance, if there are 9 seats, the threshold would be 10% of the vote plus one. Among those candidates who are not elected directly, the ones with the lowest vote totals are eliminated, and their votes transferred according to the predeclared preferences of the candidate preferred on that ballot. Once candidates are elected, each district would be assigned one representative from each winning party, so that each representative could have multiple districts; smaller, more concentrated territories for representatives of larger parties, and broader, more diffuse territories for those of smaller parties. Thus, among the representatives assigned to your district, you could always ensure that there would be one whom you'd helped elect; someone who truly represented you, both geographically and ideologically.
In order to assure voters that their delegated votes would be used in a way they support, each candidate would predeclare a simplified preference order. This means:
- Candidates choose a party, or may run as independents.
- Party candidates may choose a list of other candidates within their party to declare as "their faction". Factional affiliation need not be mutual.
- Whether or not they choose a party, they may list a preference order over the other parties and independents.
For instance: Betty Best affiliates with the Good party. Among the Good candidates, she declares Bob Better and Irma Improved as her faction (we'll call these the "Better" faction). After the Good party, she prefers (any member of) the Decent party, then independent candidate Andrew Average, then the Acceptable party. She does not support the Bad or Horrible parties.
Bob Better, as it happens, defines his faction to include Irma but not Betty. His party preferences are similar, but he ranks the acceptable party above independent Andrew Average.
If you vote for Betty, your vote will be transferred in the order she declared. That is, even though it will probably count temporarily as one of Bob's votes, it will not take on any of his preferences for other candidates.
Most voters would vote for one local candidate listed directly on their ballot. Some more politically-engaged voters would write-in a candidate from faraway. A few, who do not like the idea of delegating their vote or do not agree with the predeclared preferences of any candidate, would approve a set of several candidates; these ballots would still be counted fairly. Thus, voting is very simple, and it is nearly impossible to accidentally spoil your ballot.
Say that you vote for candidate Betty Best from the predeclaration example above. Your vote would go toward electing Best if possible; if she were eliminated as being too weak, it would be divided among the un-eliminated Better candidates; if all of them were eliminated, it would go to the remaining Good candidates; then to the Decent ones; then to Average; then to any of the Acceptable ones. If at any point in this process one of these candidates won, then whatever fraction of your was necessary to ensure that win would be used up; so probably, before all the Good candidates were eliminated, you would already have helped elect one or more candidates you'd support, and only a small fraction of your voting power would continue being transferred.
Districts would be assigned based to representatives based on their support and their party's support in that area. Thus, in an area where one party was particularly dominant, a representative from that party might cover just their local district (though that district would also be assigned as part of the territory of one representative from each of the other winning parties, so minority-party voters in that district would still have local representation from someone they'd supported). In an area where a party was weak, that party's winners would have to cover large territories.
The district assignment algorithm tries to assign each candidate the districts where their support most exceeded their average overall support. Thus, while no district assignment will be perfect, in general candidates will be sympathetic to the districts they represent; for instance, in a state where rural and urban issues were significantly different, rural districts would tend to have rural representatives, even within a party which appealed more to urban voters.
It's easy to get confused by the details of any given electoral process, because rules which cover every eventuality are rarely simple. However, this system is very simple for the voters; all you have to do is vote for your favorite candidate.
Also, it's designed to be a gentle change from a single-member-district system. The ballot format is similar; districts can remain unchanged; voters will still have a single person they can point to as "their representative"; and if single-member districts are giving fair proportions from cohesive parties, PAL representation will elect exactly the same members. Though other, older proportional systems such as STV or mixed-member have seen more use, PAL is still in some sense a safer, more gradual reform.
The difference is that most representatives will represent multiple districts, and each district will have multiple representatives (one from each winning party). This allows each voter to know who their representative is, while preserving ballot secrecy. Thus, whereas currently only 60-70% of US voters voted for their representative, and many of those because they have no real choice, with PAL voting over 80% overall, and over 95% in large states, would be guaranteed to have a representative whom they'd supported directly or indirectly.
- Thus, a large majority of voters have real representation
- Each representative is elected with the same number of votes.
- P is also Prudent; not a radical change from single-member districts
- No redistricting necessary
- all votes are for one of the two main-party candidates in the voter's district,
- all candidates approve everyone from their party
- and the districts are divided fairly so that plurality would give a proportional result
- ... then PAL representation (like Balinski's "Fair Representation") gives the same results as plurality. These assumptions will not generally be perfectly true, but they will generally be close to true, so PAL representation will give results that are recognizably similar to those of single-member districts. It is hoped that this would make it a more acceptable system to politicians who have won under single-winner rules.
- Voters, not party bureaucrats, decide which members of a given party get seated.
- Predeclared "faction" preferences give voters a clear view of and say over the ideological variations within a party, without opening the door to tricky strategic politicking as full candidate-by-candidate preference orders would.
- In particular, they are a good compromise between party discipline and individual independence; a rogue partymember can be disciplined, but only insofar as they don't have enough direct votes, and only by a natural consensus of the party, not by any centralized party bureaucracy.
- Since the total votes needed for election is higher, the "margin of victory" is reduced. There are no safe, gerrymandered seats where corrupt representatives can hide.
- Representatives know who is a constituent and voters know who is their representative.
- Neighbors can organize to lobby their shared representatives.
- Fair attention for local issues.
Compared to other PR systems
Other PR systems have problems which make them extremely hard to pass as a replacement for single-member districts. PAL resolves all of the following issues:
- A closed list system would be (rightly) attacked as a power grab by party bureaucrats. Voters have been souring on parties for decades now, and they wouldn't stand for that.
- A global open-list system such as STV would have unacceptably-complex ballots. Who can keep track of dozens of candidates, let alone fully rank them?
- A districtless system would be too radical a change. People are used to having "their" representative.
- A multimember-district system would help with the above problems, but wouldn't actually solve them. Who wants a system where ballots are only a little bit too complex, where you only sort of know who your representative is, and which is only mostly proportional?
- A mixed member system would be an ugly hybrid. US democratic ideals may be too egalitarian to accept the idea of two different kinds of representative.
- More seriously, multimember or mixed member systems would be totally unacceptable to existing incumbents, as either would draw too many of them out of their existing districts. And perhaps this is in part a valid concern. It is true that the public interest is to have representatives who are accountable, not complacent; but that does not imply that there's a value in change simply for change's sake.
- Fair Majority Voting, as used in Zurich, Switzerland for municipal elections, resolves all of the concerns above, but it would be very hard to justify the fact that some representatives would lose with a majority vote. It's very hard to respond to a simple question like "Why should my opponent win with 45%, when I lose with 52%?" with a complex answer about party balance and compensating for gerrymandering.
- Note that PAL representation would actually tend to give the same result as FMV, but would provide an easy justification for that result. Responding to the question above, you could say: "Each representative needs exactly the same number of votes to win. Your opponent got the vote transfers they needed to reach that threshold and you didn't. Those votes were transferred in accordance with the explicit will of the voters, and disallowing transfers would disenfranchise those voters."
- Candidates pre-announce their rank-ordering of the other parties. They may also designate a subset of the candidates in their own party ("their faction") as preferred.
- Voters may vote on the candidates in their or nearby districts, or write in candidates from farther off.
- To simplify the ballots, districts are grouped into sets of 2 or 3 "super-districts". The ballot for each district lists the incumbents and candidates from that individual district in a larger font, and the candidates from its super-districts below that in a smaller font. Write-ins may be used to vote for candidates from outside the super-district. Larger parties will usually run one candidate per district; smaller parties may run just one candidate per super-district.
- Any ballot which votes for one candidate only will go to that candidate as long as they remain in the running; then be divided among all "same faction/same party" candidates equally, as long as there are any; then among all "same party" candidates, as long as there are any; then among the candidates for the second preference party; then third, then fourth, etc.
- A legislature is elected by a modified version of STV, using the following steps:
- The iteration number I is set to 0. (Higher iteration numbers mean lower quotas. The process will be run with an increasing iteration number until a full slate of candidates reaches the quota)
- The quota Q is set to the rational number (V+1)/(S+I), where V is valid votes and S is seats. (For the first iteration, this is a form of the Hare quota. However, it is likely that the process will usually run for exactly two iterations, so the quota will end up being a form of the Droop quota.)
- Delegated votes first count full-weight for their chosen candidate. Once that candidate is elected or eliminated, a vote is divided equally among all non-eliminated members of approved faction of the same party. Once there are no more of those, the vote is divided among the non-eliminated members of the surviving party which was most preferred by the voted candidate.
- Undelegated votes are divided equally among all approved, non-eliminated candidates on that ballot.
- Any candidates who reach the quota are immediately and simultaneously elected, and their ballots are reweighted to eliminate a quota.
- They keep the remaining fraction of the ballots which elected them, one quota in total. This is used later during the draft phase to help decide which additional districts, besides their home district, they will represent.
- If there are no candidates who reach the quota, the party with the lowest remainder, when total party votes are divided by the quota, is identified; and the candidate from that party with the fewest votes is eliminated. All votes for that candidate are reassigned as outlined above.
- If the above finishes without electing a full slate, the process so far is recalculated with a lower quota:
- The iteration number is increased by 1. This reduces the quota Q, as if it were the Droop quota for a legislature one seat bigger.
- All elections and eliminations so far are re-applied in the same order with the new quota (see note at bottom of this section)
- The process is continued from step 5.
- For each party with more than one seat, the districts are divided among the party representatives through the following automatic "draft" procedure:
- Drafting order: The party P representative with the fewest party P votes in their assigned districts drafts. If there is a tie (such as initially), representative with the highest number of direct votes from some district drafts.
- At each step, the representative drafts the unassigned district which gave them the most direct votes.
Your representative is the member of the party you voted for who is representing your district. If no member of the party you voted for was elected, then you may look at the public preferences of your chosen candidate to see which of your district's representatives is yours.
(Note: this step used to say, rerun the election from scratch. However, that version provides a truncation incentive. This version gives the same protection from small, marginal parties as before, but removes the truncation incentive.)
Optional party threshold
Optionally, one additional rule can be added to modify step 5 above:
- No representative may be elected unless their party got at least T votes, where T is some party threshold.
This would encourage small parties to join into coalitions, and thus promote a less-fragmented legislature. There are various options for T. It could be as high as 5%, similar to the 5% threshold used in the German parliament. Or it could be as low as V/(S+I-1) (that is, if the process completes in just one iteration, the Hare quota V/S); this would actually allow independent candidates to be their own "party", but only if they have enough support to fully deserve one of the S seats.
This rule complicates the system somewhat, so it is not recommended if the PAL representation is to be implemented by a voter referendum. If the system is being passed by a legislature, they may be more concerned about fragmentation, so they could use a relatively-high 5% threshold. And if the system is implemented by a constitutional convention, a V/(S+I-1) threshold is ideally fair.
District 5 ballot
▢ ________________________(write-in; see attached list of all declared candidates for preference information)
If you only vote for one candidate, your vote is delegated, which means it may help elect that candidate's preferences. If you do not wish to delegate your vote, you may vote for more than one candidate, including write-ins.
Vote transfer procedure
Note: This example does not work well to show the district-based part of PAL representation, because there is no way to divide up the voters into three equal-sized districts; and even if there were, real-life voters would never vote this homogeneously; and the example elects only one "representative" from each "party". Thus, the example only covers the modified-STV system, and does not include the final district "draft" phase.
Imagine that Tennessee is having an election on where to locate 3 public universities. The population of Tennessee is concentrated around its four major cities, which are spread throughout the state. For this example, suppose that the entire electorate lives in these four cities, and that everyone wants to live as near as many universities as possible.
The candidate sites for the university are:
- Memphis on Wikipedia, the state's largest city, with 42% of the voters, but located far from the other cities
- Site 1
- Site 2
- Nashville on Wikipedia, with 26% of the voters, near the center of Tennessee
- The "Eastern Party", composed of:
The preferences of the voters would be divided like this:
| 42% of voters
(close to Memphis)
| 26% of voters
(close to Nashville)
| 15% of voters
(close to Chattanooga)
| 17% of voters|
(close to Knoxville)
The quota is (100.0008%/(3+1))=25.0002% (The small fraction represents one extra virtual voter, to ensure that the quota cannot be met by four different sites). Since both Memphis (site 1) and Nashville are over the quota, both are elected first. Memphis votes are multiplied by 17/42 and transferred to Memphis site 2, and Nashville votes are multiplied by 1/26 and then split evenly between Chatanooga and Knoxville. Totals are now:
- (Memphis 1: 25.0002% (elected))
- Memphis 2: ~17% (actually 16.9998%)
- (Nashville: 25.0002% (elected))
- Chatanooga: ~15.5% (actually, 15.4999%)
- Knoxville: ~17.5% (actually 17.4999%)
The party with the fewest remaining votes is the Memphis party. Within that party, Memphis 2 is the site with the fewest votes (in fact, the only remaining site), so even though it has more votes than Chatanooga, Memphis 2 is eliminated. The votes are pass over the already-elected Nashville to tranfer to the Eastern party. Within that party, Memphis disapproved Knoxville, so the full total is transferred to Chatanooga. Chatanooga now has ~32.5%, more than the 25% quota, so it is the third and final site.
If Knoxville had not joined a party with Chatanooga, then Chatanooga would have been eliminated, and Knoxville would have been the final site. But Chatanooga could have responded by threatening to prefer a second Nashville site, or even Memphis 2, over Knoxville, if Knoxville would not cooperate in the Eastern party. In the end, Knoxville's strategy may or may not have worked. In general, such strategic gamesmanship would be less profitable and more dangerous in a real election, with more seats overall as well as a significant degree of polling uncertainty.
District assignment ("draft")
Say that party P has elected 3 representatives, A, B, and C. There are 6 districts, with 1000 eligible voters each, and the following vote totals:
|District||Total valid votes||Direct votes for A||Direct votes for B||Direct votes for C||Total party P votes|
District assignment proceeds in the following order:
- A:1 (400 constituents)
- B:2 (550 constituents)
- C:5 (200 constituents)
- C:3 (200 more constituents, 400 total) Now A and C are tied for number of constituents, ...
- A:4 (201 more constituents, 601 total) ... but A goes first because they have more votes in some district.
- C:6 (200 more constituents, 600 total)
Note that C gets 3 districts where party P was weak, B gets just 1 district where party P was strong, and A has one of each. Thus, at the end, A has 601 constituents; B, 550; and C, 600. This is a more balanced arrangement than their original vote totals (771, 549, and 431 respectively).
Third example to highlight advantages
Imagine a simple state with three congressional districts and three parties called R, C, L and D (any resemblance to real party names is... purely coincidental). There is just one statewide candidate each for the C and L parties, who both declare that they prefer R. But the R candidate in district 1 is corrupt, and so is not approved by either the C or L candidates. Also, some R voters from district 1 choose to write in the R candidates from other districts rather than vote for their corrupt hometown candidate.
|District 1||60% (50% R1, 4% R2, 6% R3)||0%||10%||30%|
|District 2||35% (+4%)||5%||20%||40%|
|District 3||25% (+6%)||5%||0%||70%|
|combined candidate||10% of a district||30% of a district|
|statewide party total||40% of state||3.33% of state||10% of state||46.67% of state|
The threshold to win is 75% of one district's vote. The elimination proceeds as follows: 1. Eliminate L; votes go to R2 (44%) and R3 (36%). 2. Eliminate C, leaving R2 59% and R3 51%. 3. Eliminate D1, leaving D2 55% and D3 85%. D3 is elected; their excess 10% leave D2 at 65%. 4. Finally, eliminate R1, leaving R2 84% and R3 76%; elect R2 and R3.
After the STV process is done, the winners are R2, R3, and D3. R2 is assigned to district 2, R3 is assigned to district 3, then R3 is assigned to represent district 1; and D3 is assigned to all three districts.
Thus any voter for R, C, or L would know that their representative was the R assigned to their district, and any D voter would know that their representative was D3. Because of the unusual situation where District 1 was strong for party R, but R1 was not elected, it happens in this case that R and allied constituents are somewhat unbalanced; R3 represents 100 constituents for every 60 represented by R2.
This example shows some of the advantages of PAL representation. In district 1, voters clearly prefer party R, but their local R candidate is corrupt; even though many of them lazily vote for this local incumbent, PAL gives them an R representative who is cleaner. In district 2, party D has a plurality, but the majority is anti-D; PAL respects that anti-D majority by still electing ⅔ of the state reps from party R. And party R can't neutralize D voters by gerrymandering them into district 3; If D could get an extra 10% in any district, they'd take an extra seat. Finally, the minor parties C and L do not elect any representatives because, even combined, they have not reached the threshold of 75% of one district (25% statewide); but their concerns cannot be ignored, as either one still could hold the balance of power between R and D for one seat.
Note on legality in US
US Public Law 90-196, passed in 1967 by a voice (anonymous) vote, gave citizenship to one man. An amendment, of unrecorded sponsorship, added the following text:
- In each State entitled in the Ninety-first Congress or in any subsequent Congress thereafter to more than one Representative under an apportionment made pursuant to the provisions of subsection (a) of section 22 of the Act of June 18, 1929, entitled 'An Act to provide for apportionment of Representatives' (46 Stat. 26), as amended, there shall be established by law a number of districts equal to the number of Representatives to which such State is so entitled, and Representatives shall be elected only from districts so established, no district to elect more than one Representative (except that a State which is entitled to more than one Representative and which has in all previous elections elected its Representatives at Large may elect its Representatives at Large to the Ninety-first Congress).
Although PAL voting would usually give identical results as Fair Majority Voting (FMV), the former does not meet this law while the latter does. Thus, if PAL voting were to be passed in some state, it would have to be conditional on this law's repeal. FMV could be used in the interim.
PAL representation is inspired by Michel Balinski's Fair Majority Voting and by SODA voting. From the former, which is used for municipal elections in Belgium, it inherits the combination of geographical districts and proportionality. However, unlike Fair Representation, each candidate elected by PAL representation has received (directly or indirectly) the same number of votes. From SODA voting, PAL representation inherits the simple, spoilproof ballot format and the optional vote delegation.
A modified version of STV is used as the proportional system for simplicity. Other proportional systems might also work (although a non-LNH system might put perverse incentives on candidates). The main difference from standard STV is that this system allows equal ranking, and thus uses fractional division of votes. This is necessary for three reasons. First, it allows for approval-style votes to be counted without complicating the ballot. Second, it allows candidates to exercise judgment independently from their party (disapproving of certain party members), but keeps the voter's judgment as primary. If candidates couldn't exercise judgment, parties would have to waste energy keeping out "crazy" candidates who affiliate only because of the transfer votes they might get. If candidates could fully-rank within the party, as would happen if the PR system were standard STV, there would be too many opportunities for logrolling, at a level of detail where voters wouldn't realistically keep track or hold candidates accountable. Third, equal-ranking makes it so that, under reasonable circumstances, PAL could elect exactly the same representatives as a non-gerrymandered single-member-district system; this is an important selling point for incumbent politicians.