Maseru Observer alleges that the ruling LCD party in Lesotho manipulated the two-tier, two-vote MMP system in last Sunday’s election, setting up a shell party in the list compensation tier to increase its overall seat total dramatically. That is, by running a separate but dependent party slate that had no candidates win (or run) in the single-seat districts, it was able to win PR seats that otherwise would have gone to opposition parties. (The main opposition party apparently did something similar.)
This potential for manipulation of MMP has been recognized in the literature on Germany for as long as I have been reading electoral-system literature. However, I am not aware of a case where it had been practiced (at least on any significant scale) until an Albanian election a few years ago.
I have several readers who know a lot about MMP and/or are fans of the system. How would you prevent this sort of manipulation? One obvious way is by having a single vote, rather than separate SSD and list votes. However, that has disadvantages of its own.
(Please note that I have not independently verified Maseru Observer’s claims.)
Fruits and Votes 
Perhaps some sort of “overhang cap” would work? As a quick example, say you cap the SMD seats at 2x the proportional vote results.
Then in Lesotho for example, the LCD winning 61/79 (about 75%) of the seats would be required to win about 38% of the proportional vote to keep all those seats, or all the seats above 2x of the PR vote total get redistributed in the closest constituencies to SMD candidates from the next-in-line parties calculating proportionately. [It appears the LCD or “NIP” actually won 52% of the PR vote]. This wouldn’t really happen too often. In the extreme Lesotho example, neither major party even ran a PR slate. Therefore, ALL the SMD seats (except the one won by a small party) would be redistributed proportionately. The “shell lists” would receive the seats, which would mean a pure PR result.
I quickly ran this (with approximate numbers) for the Albanian election of 2005, in which the Democratic Party won 56/100 SMD seats but took only 7.5% of the PR vote, and the Socialists won 42/100 SMDs but won only 9% of the PR vote. After proportional redistributions of all the SMD seats above 2x the PR percentage, the seat allocation gives the Democrats and allies about 67 of 140 seats, or about 48% (and they actually took about 41% of the PR list vote). About 38/140 or 27% go to the Socialists and Social Democrats, who together took around 22% of the list vote, and LSI ends up with about 13 seats which corresponds to its percentage, smaller parties taking the rest.
Point being, it wouldn’t really get to that point, as implementing this rule would seem to end the skirting of the spirit of the law… but are there unforeseen consequences?
I wouldn’t see this being a problem in a developed democracy where it would be seen as outside the established “rules,” though…
LikeLike
Lesotho could change it’s MMP system for list PR system. The country can either use a small multi member close list PR system or 1 seat closed list PR single member districts and 25% compensatory tier. The reason for this if an MP dies or resigns in the single member district then the next person on the party list can take over without the need for expensive by-elections. When Aland which is a part of Finland votes in Finnish election, it’s elects one member in a one member proportional representation district although it is not proportionate.
LikeLike
This issue of manipulation of MMP is also being discussed over at the Head Heeb. I noted Wikipedia’s claim that such manipulation of MMP occurred in the 2001 Italian election.
Alex, I’m a little confused about your proposed fix. If you’re going to cap the constituency seats, why redistribute them according to PR? This means they lose their local representation. Wouldn’t it make more sense to have them revert to the second-place constituency candidates?
Also, this fix still leaves the system open to manipulation. For example, the LCD in Lesotho could say “in the PR section of the ballot, vote for the LCD if your name begins with these letters, or the NIP otherwise”. With perfect information, they could gain an extra 5 seats or so. As far as I can tell, this technique will only work for parties which already have an overhang (and it’s certainly not as bad in this way as pure MMP).
Probably the biggest problem with your proposal, however, is that it essentially bars independent local candidates from running. This can’t be fixed by making an exception for independents of course, since what’s to prevent a party from running all its SMD candidates as “independents”?
LikeLike
The “Berlusconi trick,” as some call it, was possible because he opposed MMP and therefore had no scruples about running a “decoy list” or “twin party” in 2001. Since it was obviously going to succeed, the left had to copy the trick, effectively turning Italian MMP into MMM.
I don’t know who invented it first, Berlusconi or Yaracuy governor Eduardo Lapi in 2000 in Venezuela. His original local party, Convergencia, ran only on the state wide list. The new party was L.A.P.Y. which improbable and coincidental name meant Lo Alcanzado Por Yaracuy (“what Yaracuy reached”, as a clear allusion to how good a governor Lapi thinks he is). LAPY ran only in the individual districts. Thus in 2000 LAPY got the 4 individual seats. Convergencia on the state wide list got 2 out of 3 and thus Lapi ruled his term with a very comfortable 6 to 1 majority. Just as in Italy, the left was forced to copy. Thus in Yaracuy 2004 Chavez’ supporters ran as MVR on the state list and PODEMOS/PPT in the individual districts. Lapi repeated his trick as in 2000, and was better at it: Lapi lost the state house but did keep a 4 to 3 majority in the legislature.
This report tells how the “morochas” (twins) loophole withstood a legal challenge.
Other than a single-vote model (which Louis Mssicotte thinks is the best way to prevent this trick) I know of two ways to stop it. One, a strong Election Act giving the Chief Elections Officer power to rule that the twin parties are really one party. Two, a Minimum Count Provision. A party’s Final Total Vote shall be the lesser of its Initial Total Vote, or, two times the sum of votes received by its local riding candidates. This provision, to take an example, would not penalize a party getting 12% of the party vote unless they had less than 6% of the local votes.
This makes regional candidates somewhat dependent on the success of their local counterparts. This encourages cooperation between local and regional candidates of a party (if they are not already dual candidates), who can otherwise have an incentive in MMP systems to hinder each other because of the compensating mechanism between regional and local seats: the fewer local MPPs the party elects, the more regional MPPs it gets.
LikeLike
In Germany, if independents or candidates from parties below the thresholds are elected to single member seats, those voters who supported them have their list votes excluded when list seats are calculated (see the two PDS constituencies in 2002 for an example). Of course, in practice the exclusion only affects those who split their votes. This process is only possible when the two votes are traceable to each other, for instance by using just one ballot paper as in Germany. The system does not remove the possibility of a party still receiving some mathematical advantage by intentionally de-linking its constituency and list candidates, but it reduces it greatly.
More importantly though, this system does not cover overhang seats, which similarly can be used to cheat the system (so long as the constituency winners are tied to lists that passes one of the thresholds – here the three-seat hurdle could easily be overcome). It is possible, however, to expand the system by devising a formula that would seek to penalise related vote-splitting at a rate up to equal to any advantages resulting from overhang seats. For example, if the CDU won overhang seats in Baden-Württemberg partly due to the fact that some voters used their first vote to support a CDU candidate and their second vote to support, say, the FDP list (this happens quite often), that would entail excluding a proportionate amount of FDP or other list votes in a final recalculation of list seats. There might be a slight problem if the cost of a list seat is not preserved at the previous level. Why not just exclude all ballots that supported these constituency winners? That might punish a list too hard.
There is a simpler but cruder method, which I may have concocted myself. Unlike the above system, it does not pulverise the transfer of penalties onto multiple lists. When overhang seats and the like appear at the counting stage, assign such seats artificially to other seat-eligible lists, based on relationships between how the two votes were cast. Note, the constituency winner remains the same, the assignment is for list seat calculation purposes only. The process can be done in two ways: Firstly, by ranking the rate at which those who voted for the different lists simultaneously voted for the constituency winner. Based on this percentage, a list might then find that a seat belonging to another party is being counted as if it were its own. The second way would be somewhat biased against the larger parties, by simply ranking the number of list votes among those who supported the constituency winner. In whichever direction the analysis is done, if a party had eight constituency winners and three overhang seats, each constituency analysis would determine the fate of a subtraction worth 3/8 of a seat. Following the above example, the CDU would then keep its overhang seats, but entirely at the expense of lists whose voters to a large extent supported CDU constituency candidates. If limits are set and/or the vote analysis does not produce any such patsy or clone lists, the CDU would as now keep its advantage, due to the fact that no significant strategic or split voting was detected.
Finally, it is worth mentioning that overhang seats usually do not appear because parties or voters intentionally try to squeeze unfair advantages out of the cracks in a system. Therefore it might seem unfair to penalize their friends. Systems that seek perfection may also be perceived as abhorrently complex. Still, as the Lesotho example shows, it is worth having a system in place which at least reduces the likelihood of being rewarded for “fraudâ€, which in turn encourages straightforward behaviour.
Now breathe.
LikeLike
Vote-splitting in MMP is not cheating at all. The very purpose of the two-vote system is to allow vote-splitting, so that you can reward or punish your local MP without hurting your preferred party.
What is cheating, of course, is the “twin parties” trick which Germany has never been plagued with.
Less obviously cheating is the tactic of the Scottish Green Party: they run no local candidates, claiming this is a waste of scarce resources. But what they are really doing is saying to Labour Party voters “Labour wins so many local seats that your party (regional) vote for Labour is wasted, so we’re letting you win the local seats, and you can reciprocate: “give us your second vote.” This suggests that they can make the Greens their second preference. This of course is not how MMP works: the party (regional) vote is the vote that determines the make-up of parliament. This confusion will be all the greater when local elections are held under STV, where “second preference” means just that.
Accordingly the ballot for the coming Scottish Parliament election will be changed as recommended by the Arbuthnott Commission. The British government announced on 22 November 2006 that the two separate ballot papers used in previous Scottish Parliament elections would be replaced for the elections in May 2007 by a single paper—with the left side listing the parties standing for election as regional MSPs and the right side the candidates standing as constituency MSPs. The Greens will have to find a new slogan.
LikeLike
Is the “Berlusconi trick” relevant to this discussion? Italy’s mixed-member system was not MMP, the classification as such by some reputable sources (and even Wikipedia!) notwithstanding.
LikeLike
Would not the easiest solution simply be not to permit a party to register a list unless it has a minimum number of district candidates?
I am sure there is a downside; there almost always is. But I do not like any solution that penalizes split voting, per se, just one that penalizes parties setting up an “evil twin.”
I think the Minimum Count, mentioned by Wilf, is also potentially valuable.
I agree with Espen’s concern about avoiding unnecessary complexity, which can undermine the support for the system, and also often introduce unforeseen other “manipulation” tactics.
Thanks to all of those who have weighed in. It is what this blog is all about! And please keep the ideas sprouting!
LikeLike
The situation in Italy is certainly relevant. The twin-parties strategy works with most mixed-member compensatory systems. And although Italy is not “classical MMP”, it certainly is both mixed-member and compensatory (if only modestly so). So why shouldn’t it be considered?
I like Wilf’s local-vote threshold for PR seats. My one concern is the difficulty it poses for small, spread-out parties. One likes to think that in a PR election campaign, every party, no matter how small it was previously, has a chance to earn representation if their issues become important to voters (see the Pensioners’ Party in Israel for an example). With a local-vote threshold, very small parties that can’t afford to run many constituency candidates give up any chance of a surprisingly good result–they’re just running for show.
Perhaps changing the Total Vote formula could help. Instead of Vtot = min(V, 2 * Vconstituency), maybe something like Vtot = min(V, 2 * Vconstituency + V(S)), where V(S) is the number of votes required for some predetermined very small number S of PR seats. This means that a party without local candidates can still reasonably gain representation, but limited to S seats. It is definitely more open to manipulation, but to cheat effectively you’d have to set up a multitude of front parties. Better ideas, anyone?
LikeLike
Just to clarify, my proposals would not penalise split voting, except in the relatively few cases that resulted in overhang seats or in the election of candidates unattached to any list eligible for seats.
Split voting ordinarily is not cheating, but attempts at cheating most often involve forms of split voting. It is not the only strategy for achieving unfair advantages, but it is perhaps one of the most difficult to counter because it involves how and why candidates are entered and how and why voters choose among them.
I expect objections against the definition, but the Italian and Lesotho cases are essentially examples of forced, party-organised vote-splitting on a massive scale, with entirely predictable rewards. Also, when someone attempts to give extra weight to his or her votes through tactical CDU/FDP splitting at federal elections in Baden-Württemberg or Labour/Green splitting in some regions of Scotland, this may in itself be much less effective, but it has only slightly less predictable potential rewards.
No doubt it is possible to avoid much abuse indirectly through many of the models and requirements you all have proposed, though complexity will probably still be the price of robustness. My own attempts were instead aimed at attacking directly a problematic cause of disproportional advantages and of cheating incentives, and to do so without really interfering much with how the various political actors naturally would choose to organise candidates and votes to their maximum benefit, in the single-member seats, and with the lists, respectively.
LikeLike
Vasi, the Italian case was designed to be majoritarian, with “modest” compensation. Whether that makes the case more or less relevant to the gaming of MMP systems is a matter of taste, I suppose.
I share Vasi’s concern about the Minimum Count Rule’s potentially deleterious effect on very dispersed small parties that might not obtain many SSD votes, even if they have the candidates. (And, really, if they can present a list, they can present the same candidates in the districts, as long as the nomination procedure is not onerous.)
Espen, thanks for your clarifications.
Obviously, this is not a simple matter!
LikeLike
One idea I have for a one vote MMP system is to allow independents and candidates not choosing to affliate with a party list is to put all Independents on a Assorted Party List and the Independent Electoral Commision randomly chooses those the order of those people on the list by lot. That means treat all the independent candidates as a party. The same could be adapted for a two-vote MMP too.
LikeLike
In Lesotho at least, two rules would have resolved the problem. (1) All parties fielding constituency candidates are required to be on the party list. There was apparently a lot of confusion about LCD not being on the party ballot. This would have made it a lot more difficult to explain to the mostly uninformed voters why they should vote for one party on one list and another on the second. (2) An individual cannot stand for one party in a constituency, and be on the party list for another party. This seems self-evident, and it is surprising if Lesotho does not have this rule already. What is the prime minister doing on the party list for a fringe party??
LikeLike
There has been some discussion about ‘cheating’ MMP in Wales as well. Constituency level Independents would not count against the entitlement of list Independents, a loophole that may be exploited by former Labour Minister Ron Davies, now standing as an Independent in a constituency. An ally is standing as an Independent on the list, and in that region there is another seat which currently has an Independent AM (though she is not affiliated with Davies).
There are some concerns about this being a dodge to get round the rules, given that Davies has previously been involved in a party called Forward Wales and that this may in effect be an organising entity behind the Independents.
The Welsh electoral system (unlike that in Scotland) prevents ‘dual candidacy’ (after a divisive argument on the issue in 2005-06). Some perceive the Independents’ scheme as being a justifiable response to a previous change in the system that disadvantaged smaller parties.
In the AV+ system recommended by Jenkins for the UK House of Commons in 1998, parties putting forward a list would be required to stand candidates in at least half the component districts, which might make the dummy list too complicated a trick to pull, particularly given a situation where there are legal requirements on registered parties.
LikeLike
What is an “independent”? I am confused by the way both Suaprazzodi and Lewis are using the term. As I understand the term, an independent is some candidate with no party affiliation, by definition. That means no affiliation across independents, as well as no affiliation of individual independents with candidates of other registered parties. So, how can you have a list of independents? As soon as you have a list, then you have a set of candidates sharing a common electoral-allocation fate. In other words, a party.
LikeLike
That is what I mean is to pool all the independent candidates as a party on a Assorted Party List. The electoral commission decides by lot which independent will be first on the list and so on.
LikeLike
In answer to Matthew at 15, in the UK MMP systems an individual can stand in the regional list part of the election as essentially a list of one – I am not sure whether they can group together without becoming defined as a party, but I doubt it in that context. However, in some local authorities in the UK ‘Independents’ form collective Groups, and sometimes there are more than one such on the same authority, or one Independent Group and some maverick double-Independents!
In Scotland in 2003 an Independent candidate, Margo MacDonald, won a list seat. If such a seat falls vacant, it remains vacant because there is nobody further down the list. In the Welsh case if Independents win a regional seat it would be one at most, so there is no problem with that unless the elected candidate dies or resigns. The question of what degree of co-ordination between Independents is permissible without being a party is one that might exercise our Electoral Commission in due course.
LikeLike
Maybe a controversial solution, but definitely an effecive way to end ‘decoy lists’ being used, would be an MMP system with the constituency vote used to calculate the overall proportional result. This way vote-splitting would also be impossible, there only being one vote.
LikeLike
Germany I think has a provision, (maybe at Lander and local level?) for associations of electors to put forward non-partisan lists. I suppose this isnt exactly an independent by Anglo-Saxon definition, but it is specifically non-party.
Another example I can think of were the closed list election for the Northern Ireland Forum in the late 1990s. The districts were coterminous with the Westminister constituencies, and I guess one had to have complete lists, because there were a handful of “Independent John Smith” lists, presumably close friends and family obliged their ambitious/principled relative to fill out the vacant places.
LikeLike
JD, the very first election in West Germany had one-vote MMP.
Of course, banning split-ticket voting means many of the advantages of MMP are thereby lost. For example, voters who favor a smaller party are unable to indicate a preference among the major contenders for their district representative.
I would say that allowing only one vote for both local candidate and party list is a cure worse than the disease, given that there are other ways to address the problem of “decoy lists”, as others have noted.
(Thanks, JD, for bringing this old discussion back to life!)
LikeLike
What about keeping two votes but allocating list seats in proportion to each party’s total of both – ie, its local votes for district candidates plus its Zweitstimmen for the list itself?
LikeLike
MSS: It is indeed a bit or a bitter pill, which is why I called it a controversial solution. Interestingly enough however, the last NZ elections suggest to me that it might not be a big issue, or at least that most NZ voters wouldn’t see this as a problem. About 7% still voted Green for their constituencies, and there was a higher constituency vote than list vote for ACT. NZFirst had a much lower constituency vote to list vote ratio, which is probably due to the small number of constituency candidates they fielded. Lastly, the National vote was the same percentage for both votes. (source:http://psephos.adam-carr.net/countries/n/newzealand/newzealand2011.txt)
LikeLike
A drastic solution that could be implemented would be to say that any one who wins a constituancy, who does not belong to a party that has crossed the electoral threshold, would have very restricted voting rights where they can not vote on the budget or on legislation except private bills that only effect people within there ridings. Bills that have to do with public policy such as legalizing gay marriage would be restricted to those members who belong to parties that have crossed the threshold. Although independant and small parties could still represent the people within there riding by raising questions during question period, introduce bills and belong to committies, they would not be able to do the things that could push a vote one way or the other which is essentially what a party would want to implement its agenda. By doing this, it would eliminate the incentive of adopting this tactic.
LikeLike
Actually JD’s solution may not be that bad. One thing you could do, is to adopt an approval voting method where people could vote for as many parties/candidates as they wish. This way, a voter could cast a vote for their preferred party, in addition to an independant or small party candidate that they also like.
LikeLike
Pingback: Lesotho election, 2015 | Fruits and Votes
Pingback: South Korea 2020 | Fruits and Votes
Pingback: Is Scottish MMP being “gamed”? | Fruits and Votes