Germany: A change to overhang provision in MMP?

Is Germany about to revoke or modify its provision on “overhang” seats? Evidently there has been a Constitutional Court finding today against the current practice, ((I vaguely recall someone might have mentioned this case in another thread here.)) and there is now a debate about how to respond.

All I know at this point is a not-very-detailed story from The Montreal Gazette that Rob Richie of The Center for Voting and Democracy sent me. (Thanks, Rob!) Here are some excerpts, with the usual journalese about “complex” electoral systems.

Germany’s highest court declared the country’s complex electoral law unconstitutional Wednesday and ordered for it to be overhauled before the next general election…

The current voting system was passed only last year by Chancellor Angela Merkel’s governing coalition in an attempt to satisfy previous criticism by the Federal constitutional Court.

…the constitutional court again criticized that parties which win more seats than they would take under a purely proportional system can keep those seats — potentially skewing election results.

Merkel’s Christian Democrats and its Bavarian sister party gained an additional 24 seats this way in the 2009 general elections.

The provision is actually–dare I say it?–more complex than that. It is pretty much impossible under MMP not to allow a party that wins more than its proportional share to keep some if its resulting advantage. The only sure way would be to revoke a seat it had actually won in a single-seat district.

The German MMP (as also in New Zealand) adds further seats to the legislative chamber to “balance” these overhangs. These partially compensate other parties for the fact that some party is over-represented from its winning more single-seat districts than its proportional entitlement would be (based on its list vote, at the state level in Germany). It is not clear from the news story exactly which part of this process has been declared constitutionally invalid.

It certainly is the case that, even with the balance seats, the presence of overhangs means a “skewing” of results away from strict proportionality. Indeed, if one does not want this, one should use pure PR and not MMP. The potential over-representation for a party that performs especially well in single-seat districts is one of the ways in which MMP is a “mixed” or hybrid system, and not simply a proportional system.

84 thoughts on “Germany: A change to overhang provision in MMP?

  1. @MSS, Vasi,
    The DPR Voting process is similar to MMP and counts the votes to achieve similar ends.
    The vote for the Party confers proportionality on the whole.
    The vote for the Individual constituency MP elects the MP but does not affect the parliamentary party’s votes.

    It is of course not ‘mixed member’ because all members are elected as constituency MPs.

    With both DPR Voting the overall Party parliamentary votes are proportional, but with DPR Voting there is no overhang problem.
    Because with DPR Voting all members are constituency MPs, it deals with the problems in MMP associated with mixed members.

    DPR Voting has not been formally validated (or invalidated). How are electoral systems validated?


  2. @Alan,
    These are two important aspects.

    The 2010 Vote: As the sole representative of the Green Party in the parliament, Adam Bandt has a heavy vote conferred on him by the Green Party, and thus exercises their total vote (11.7 votes). How this works out in practice would be a matter for him and the Green Party. However it does mean that the Green Party has a vote in the parliament proportional to its election vote.

    The 2007 Vote: You are correct that in DPR Voting a small party could end up with no representation. However it is difficult to translate votes from one election to predict the result of an election under a different system.
    Campaigning and voting would be rather different. Eg There would be more constituency seats contested than in a comparable MMP election.

    Small parties would put much more emphasis on winning at least one constituency seat.
    In the UK smaller parties are relatively more effective at winning constituency seats, despite the fact that the vote for the party is conflated with the vote for the individual.

    The system also allows for the election of a single party MP with a single vote if the party achieves an agreed threshold, but fails to win a constituency.

    Nevertheless, DPR Voting would tend to limit the number of small parties present in the parliament.


  3. I suggest that large parties would carpet bomb any electorate where there was the remotest chance of a minor party winning. I am interested in the rationale for the punitive measures against independents and minor parties. It seems to me a fundamental flaw if a discrimination against independents and minor parties is actually designed into a system and I wonder why you have done it.

    My first impression is that this would be a gift to the major parties.


  4. @Alan 56
    ‘ ….punitive measures against independents and minor parties.’
    DPR Voting is designed to replace FPTP as used to elect the House of Commons in the UK, while retaining the single member constituency system.

    Compared with FPTP, Independents and small parties may expect to do better in a DPR Voting election because the vote for the individual is no longer conflated with the vote for the party. This also works for small parties if they can put up at least some well respected / popular / hard working / charismatic / local candidates.

    MMP might be considered an alternative, but DPR Voting means no change to the existing constituencies and no List MPs.


  5. @Stephen

    I accept all that. What I am asking is why you need to distinguish independents and minor parties as you do. As I understand it an MP’s vote is V/M where V is the votes received by the party and M is the number of members elected by the party. In fact your example gives the case of Party E and F who receive identical popular votes but E gets 17.5 parliamentary votes and F gets 1. Why? I realise one wins a constituency and one does not.

    Further I think it is radically unfair to clump the independents into a confected party. There are 3 official independents in the House of representatives. 2 are in broad agreement on most issues. The third takes very different policy positions and in fact gave his confidence to the Coalition rather than Labor. Why should independents not simply get the same formula as everyone else?


  6. @Alan, (part 1)
    An important part of this electoral system is that the individual MPs are elected as a result of election contests in small local constituencies. The smallest geographical area is best because the electorate has the best chance of knowing, or knowing about the candidates most directly, hence the single member constituency.

    ‘Automatic election’ allows for a party that fails to get a single constituency representative to have some representation in the House, but not a proportional vote, only a single vote.

    I understand that this appears arbitrary, which it is. It could be decided differently.

    However a consequence is that it is important for each party to contest the local constituency elections. If the candidate elected by ‘Automatic election’ had the same heavy vote as the single directly elected candidate (as per the example) it would diminish the importance for small parties of contesting the constituency elections.

    I think for democratic reasons, the candidates of every party should be examined by the election process.
    To a lesser extent political parties that have a certain level of popular support (the threshold) should be represented in the parliament, and thus have their party and policies examined more carefully than would otherwise be the case.


  7. @Alan 58, part 2
    I am worried that I haven’t grasped your point.
    Independent candidates are elected as individuals without party allegiance. There is no party vote for them to share. By definition their vote is always ‘non party political’ so they have a (default) vote value of one.

    This also applies to party sponsored MPs when they vote on any issues that are agreed by all the parties to be ‘non party political’. They have a vote value of one. This seems to me to be both logical and consistent.


  8. I am assuming that independents can appear on both the party and the representative ballots. If not, why not? Indeed what is the purpose of the party ballot except to ensure the existing Big Three in the UK remain the Big Three?Seriously you seem to have included a number of arbitrary devices whose only effect is to entrench the existing major parties at all costs. Admitting their arbitrariness is not an argument in their favour.

    On the broader point, elections are not about parties or candidates, they are about electors. You propose a system where the electors of Party E are counted as being worth 17.5 times the electors of Party F. That is a fairly radical step and needs a great deal more justification than you have provided.


  9. > “The smallest geographical area is best because the electorate has the best chance of knowing, or knowing about the candidates most directly, hence the single member constituency.”

    I dispute this very strongly A number of times as a constituent I would have much preferred to approach an MP who was not “my” “local” MP – whether because of the former’s party allegiance, or their greater expertise, or simply because I knew them personally. I have moved house a lot so I find electoral boundaries quite arbitrary. In one case I lived at the far end of one or another electoral district and was much nearer the the office of an adjoining district’s MP than to my own district’s.

    Having handed out how to vote material at polling booths I have noted that numerous voters want to vote for a particular candidate of their own party who has a high profile in the region (usually due to hard work) and get quite annoyed when told that they have to vote for someone they have never heard of (or from) because of a line on a map.

    It is true that STV uses electoral districts as well but, since these cover whole regions rather than dividing them, the above problems are much less likely to occur.

    To my mind, the fact that MMP preserves single-member districts is a bug, not a feature – especially when their role is rather like caged animals in a zoo because it’s the party list that really counts.


  10. @Alan 61
    In DPR Voting, only qualifying parties appear on the party ballot. Independents are individuals, not parties, and do not appear on the Party ballot. They only compete in the constituency contest where the description on the ballot paper is ‘Independent’.
    The purpose of the party ballot is to measure the support for each party, much like MMP, This is then used to share out the parliamentary votes in a way proportional to the number of votes each party gets in the election, which is fairer than FPTP.

    I think it is dangerous to draw firm conclusions about how voters will exercise their votes when the electoral system is changed. DPR Voting will tend to restrict the number of parties present in the parliament, depending on where the vote threshold is set (and how the party qualification rules are drawn up). However it could make it easier for small parties to be represented in parliament, and for independents to be elected.

    For Parties there are in effect three thresholds – Party qualification, the threshold level of votes, and the election of one constituency candidate. Any party who fails to pass one of these thresholds will be disadvantaged compared to others that succeed. Thresholds can be considered arbitrary, but serve a practical purpose and can be modified in the light of experience.


  11. @Stephen

    I am forced again to ask ‘Why?’ Merely restating an arbitrary design feature, especially one that can make for massive inequalities between electors, is not an argument for it.


  12. @Tom Round 62
    The proposal that constituency boundaries should ideally follow the boundaries of natural communities seems key to your argument.

    Most people would agree with this proposition, but often boundaries and communities do not coincide for reasons of the overall fairness of the electoral system (the attempt to reflect the relative strengths of the different parties), or political advantage (gerrymandering).

    A feature of DPR Voting is that constituency boundaries can be redrawn at will without affecting the outcome of the Party vote, so boundaries could be redrawn to follow natural communities of different sizes with only limited implications for Party political fairness.

    If boundaries are drawn so that they are logical in respect of natural communities, are smaller constituencies more democratic than larger constituencies?
    The smaller the constituency, the closer the electorate can be to the candidates. That is, closer both geographically, and in terms of their knowledge of the candidates, and this seems to me a democratic positive.

    Is a multimember election more democratic than a single member election?
    A multimember election puts more demands on the electorate to absorb more information about the larger number of candidates in order to make an informed choice.

    I suppose your attitude to MMP, DPR Voting or STV depends on your concept of the elected representative. Do you prefer one single local representative for a smaller geographical area, or do you prefer a looser arrangement – MMP and the additional List MPs, or STV and the choice of several MPs covering a larger area? I think there is much to be said for making politics as local as possible.


  13. @Alan 64
    DPR Voting combines electing members in single member constituency elections with a separate poll of party support. The system, as designed, encourages every small party to compete in at least one constituency election.

    If achieving the threshold level of Party votes gave the same representation regardless of whether or not the party competed in any constituencies, there would be no incentive for small parties to compete in any constituency elections, as can occur in some MMP elections.

    In my view this would undermine the significance of the local constituency election, and weaken the credentials of the system as a democratic alternative to FPTP.

    As designed, DPR Voting is intended to replace FPTP as used in the UK to elect the House of Commons. It can be compared with but is not intended as a replacement for MMP.


  14. Maybe this intervention comes far too late, but I really would like to see the thread remain (mostly) focused on the issues raised by the German constitutional court, and how MMP systems deal with overhangs and compensation.


  15. Alan (41),

    My argument has nothing to do with preferential voting. Our Constitution requires senator and MHRs to be “directly chosen by the people”. Thus, any voting system that prevents the people making a direct choice of individual candidates is unconstitutional. A closed list prevents them making a direct choice and is thus unconstitutional. First past the movable post, preferential and STV (even with the option of above the line voting) allow a direct choice and are thus constitutional. The Constitution was written when first past the movable post was the standard voting system in the colonies and must be read as accepting that system, poor though it is.

    Tom (42),

    Above the line voting is simply one of many millions of orders a voter can choose to make a direct choice of candidates and is thus, I believe, constitutional. If the option of voting below the line were removed, it would then become unconstitutional.

    I think the constitutional position with open lists is uncertain, but out of respect for the host’s request (67), I will leave the argument for another occasion.


  16. Well then, to interrupt with boring old German MMP, in reply to Rici from the weekend (no. 40):

    The representation of the East was certainly skewed at the 2002 election, but on the other hand over-representing the other parties there would not necessarily be better for those PDS voters. I think this is mostly a failure of the alternative threshold to recognise regional voices that perhaps ought to be represented.

    On overhangs I agree with you that they should be baked in during the national calculation, but I would not deprive other Land lists of any whole Hare quotas they have earned (this would have been more elegant in old Hare/Niemeyer days). If this proved inadequate, one could also add Ausgleichsmandate.

    A potential problem with the latter is if small, geographically concentrated regional parties got overhangs (or as in NZ, a Maori Party). In an extreme scenario hundreds of extra seats could then be needed just to offset a few. The opposite danger (i.e. letting associated Land lists take much of the hit for overhangs, but not guaranteeing national proportionality), is that parties would formally decouple some of their strongest areas. As long as three constituency seats were certain and overhangs were possible, fake CSU parties could be created with little mathematical risk.


  17. Wouldn’t Germany’s issue have been solved simply by calculating all seat allocations at the national level by party, with additional seats added for overhangs?

    This party amount would be fixed then. And then allocations would be made based on national party vote. If this results in an overhang for one party in one state, the process is iterative, and simply reduces the other’s states allocation for that party until all of the party’s fixed national total are allocated.

    This seems incredibly simpler and would effectively reduce overhangs since the overhangs would be adjusted more easily at the national level.

    I think the reason that this has been done this way was due to Christian Democrats’ overhangs in certain states.


  18. Espen,

    I agree that it would be odd to compensate other parties in the East in 2002 for the failure of the PDS to reach the threshold. But I don’t see this as a failure of the alternative threshold; rather it shows that the 5% threshold is far too high. The PDS achieved a national total of 3.99% in that election, and that should have been sufficient to give them list allocations.

    Similarly, in 1990, Grüne failed to reach the threshold, receiving only 3.8% [fn1], and the West was consequently underrepresented (slightly), You can see this peak also on the graph in my spreadsheet (Choose the sheet “Summary” rather than “Data”.)

    So it seems to me that setting the threshold to two or three percent instead of five would improve both district and party proportionality, and enfranchise more voters. As I said before, that would be my first reform.

    As to overhangs, I think computing the overhang nationally would suffice in pretty much all cases. The only question is how to actually do this. I think this is more difficult with Hare/Niemeyer than it would be with Sainte-Laguë (which otherwise produce very similar results).

    I probably wasn’t very clear in my earlier post (and I don’t think I’d thought it through sufficiently). Here’s a possible allocation algorithm: First compute the total national allocation per party, as per the pre-reform system. (I’d use Sainte-Laguë here, too, but it doesn’t make much difference). Then, for each party, reserve the directly-elected seats in each Land and proceed with the Sainte-Laguë algorithm until the entire allocation computed in the first step is distributed. (That is, while working down through the quotients, if you hit a pre-reserved seat, just skip it and go on to the next quotient; otherwise assign it and reduce the number of seats to allocate by one.)

    That will have the result that in Länder with an “overhang”, the party will receive no list seats, and in a few other Länder they will receive fewer list seats (probably one fewer) than they otherwise would have won. The allocation of other parties will not be affected at all. This does not solve the district over-representation problem, but in practice I think that problem is not very drastic. I’ll try to do some more simulations when I get another free moment.

    I definitely think adding compensatory overhang seats (or however you want to call these) for other parties would be a mistake. As you say, it has the potential to blow up the parliament to gigantic proportions, in the event that the overhang belongs to a small regionally-concentrated small party (or pseudo-party created for the purpose).

    Germany does not appear to have such parties at the moment (unless you consider Die Linke to be one [fn2]) but it seems that they can show up at any time.

    fn1. That only tells part of the story. In 1990, the recently incorporated East had a separate threshold, and in the East the Greens ran as part of the Green/B90 alliance which did win representation, although their national vote was only 1.2%. In the West, Grüne reached 4.7%, and all of those voters were effectively disenfranchised. The two lists combined would have just made the threshold, by my reckoning. That same year, the PDS achieved representation with only 2.4% of the vote, due to the split threshold (I think), and in the next election it made the threshold by electing four MPs from the nominal list, although it only reached 4.4% of the vote.

    fn2. It seems to me that the PDS and Die Linke are regional parties in the sense that the NDP is a regional party in Canada and not in the sense that the Bloc Québécois is; they are not parties which pursue a pro-region agenda, but rather parties whose support tends to be in certain regions. The distinction is not arbitrary, because in the latter case, the pattern of support is much less fixed, as the Canadian example shows.

    Chris (and others),

    Please don’t feel that you cannot continue your discussion. Just choose a more relevant thread to continue it in. AIUI, F&V threads are never closed, precisely for this reason.


  19. Rici and Nick,

    My quarrel on overhangs is when these come entirely at the expense of the other Land lists of a party. This is fine up to a certain point, but if the internal geographical distribution becomes too predictably skewed, it creates a win-win incentive to de-link some lists for partisan and regional gain*. I am not a big fan of federal Ausgleichsmandate, but here they would certainly remove the partisan incentive.

    The simplest way to do my calculation is to first determine which is higher for each Land list: The number of associated constituency seats, or whole Hare quotas (which now having done the math, I will amend to one 598th of all valid and invalid votes). Added up, these minimum party totals could then create overhangs in light of the Sainte-Laguë calculation of 598 seats (whereupon Ausgleichsmandate could potentially also be created). Finally, geographically unallocated seats would be distributed using basically the same method as rici’s.

    In 2009, your proposals would have resulted in a Bundestag of 601 seats. The CSU would have kept their three overhangs, whereas the CDU would have kept none, for a total of 173 seats. Since the latter also happened to win 173 constituency seats, there would have been no seats left over for its Land lists. In Brandenburg the party would have gone from five seats to one, in Bremen from one to zero. However, by strategically de-linking Land lists, the CDU could have optimally recouped about as many seats as they now have overhangs (two thirds of which in two Länder only).

    My proposal would have resulted in a Bundestag of 613 seats, and removed nine overhangs from the CDU. These would have been kept at the exact limit set by the court – not to be mistaken for absolute proof of constitutionality (though the 2009 election was the most skewed one yet in terms of overhangs). Ausgleichsmandate would have added another 35 seats for a total of 648, but valid methods limiting these could probably be devised.

    * This is not entirely farfetched: The two areas with the most secure overhangs (and three-seat minimums) within the CDU, Baden-Württemberg and Sachsen, both have long-standing non-Prussian regional identities. Also, it is doubtful if name changes would need to go beyond “CDU B-W” or if contacts with the mother party would need to be greatly affected.


  20. Espen,

    Just so I get your method your formula for seats to be awarded to a party within a state would be



    M=mandates won by a party in one state
    s=saint lague allocation based on percentage of national vote of national saint lague allocation for the party based on its national vote
    f-SMDs won in the state
    h=national hare quotas (adjusted to include spoilt votes) won in that region by the party

    and where s or f is greater than s, additional augsleichmandate would be added.

    One question, you included spoilt votes, what about votes for below threshold parties (I’m assuming right now, those would be included as well).

    Please confirm if I’ve screwed anyhing up. Now to my response rant 🙂

    Any two-vote MMP system (and even one vote MMP systems) already provides significant incentives to de-link parties in regions where a party has significant chances at winning SMDs. It would not be a national allocation system that would provide much more incentive, would it?

    Germany is basically begging for a decoy list. And some might argue, the separate existence of the CSU would be such an example.

    Really your issue is with MMP. MMP will always provide a de-linking incentive for parties likely to win SMDs.

    A system which would actually neutralize that incentive would be one-vote “scorporo” (as previously used in the Italian senate).

    Moreso, your example only met the court’s threshold of 15 seats at the exact threshold. What if the next election is more wonky?

    Finally, your proposal is fundamentally less proportional by increasing the overhang count. I have not yet calced the Gallagher Index on it. But I’d assume the Rici and I’s suggested approach would be much closer to a desired proportional outcome at the party level.

    You’re basically increasing pluralty bonus to ensure the regions which voted less for the plurality party are represented in the government.

    BTW, could someone explain to me how the court putting a cap on extra extra seats is more proportional? Do they realize that awarding extra seats actually results in more proportionality?


  21. Sorry, I think I need to revise my understanding of the formula.


    M=mandates won by a party in one state
    s=saint lague allocation at state level based on percentage of national vote (but adjusted for SMDs on an iterative basis) of national saint lague allocation (inlcuding overhangs at national level if needed)
    f-SMDs won in the state
    h=national hare quotas (adjusted to include spoilt votes) won in that region by the party

    Sorry again please confirm.


  22. Nick,

    Let me instead go through each step of my proposal more thoroughly:
    1. Current thresholds are respected.
    2. Take that largest number for each Land list (either constituency seats won by associated candidates, or whole Hare quotas based on list votes). Add them up nationally for each party.
    3. Calculate 598 seats nationally using Sainte-Laguë.
    4. If a party receives more seats nationally under point 2 than under point 3, they keep these as overhangs.
    5. I did not commit to Ausgleichsmandate since they may not be constitutionally necessary.
    If added, they would be calculated nationally to offset overhangs created under point 4. This is easily done by using the national vote/seat ratio that is lowest among the parties and applying it as a quota to the others (with normal rounding). It may be constitutional to find ways to limit the number of such seats.
    6. If a party receives more seats under point 3 (and 5) than under point 2, these are distributed among its Land lists using Sainte-Laguë (taking into account minimum seats under point 2).

    I use all valid and invalid votes as a basis for my national Hare quota under point 2. This is however just a pragmatic way of ensuring that parties keep most of the Land list seats that they have “earned”, or could have recouped had they de-linked lists near-optimally, while somewhat limiting their number so that more constituency seat overhangs can be absorbed during the national calculation.

    Now to your fine “rant”:

    “[MMP] already provides significant incentives to de-link parties in regions where a party has significant chances at winning SMDs. It would not be a national allocation system that would provide much more incentive, would it?”

    Let us leave out issues of constituency candidates having incentives to de-link from their Land lists, since none of our proposals deal with that. Currently, German MMP does not however give an incentive to de-link one Land list from another, since overhangs are all added to the total of each list. Your national allocation system would create such an incentive to de-link, whereas mine would only do so in a negligible way.

    “Really your issue is with MMP.”

    Possibly true in one sense, since letting Hare quotas based on list votes create overhangs is arguably in violation of basic definitions of MMP. Generally, I tend to favour small adjustments that limit perverse incentives.

    “[Y]our example only met the court’s threshold of 15 seats at the exact threshold. What if the next election is more wonky?”

    The 2009 election was the most severe test so far, but hypothetical courts could well strike down my system. I would then rather add a few Ausgleichsmandate than further cut list seats.

    My proposal would however have prevented ALL overhangs until 2009. If instead my national Hare quota kept using only above-hurdle votes as a basis, some overhangs would have remained: In 1994 six for the CDU, in 1998 seven for the SPD and in 2005 four for the SPD. In 2009, six more would have been added to the CDU.

    “[Y]our proposal is fundamentally less proportional by increasing the overhang count.”

    Compared to yours this is correct, unless parties started strategically de-linking their lists due to the new incentive to do so. Compared to the current system (before and after the 2011 reform) it is more proportional. I never really had a problem with the minor disproportionality caused by overhangs. But then the court deemed negative vote values for parties to be unconstitutional, which seemed to throw the validity of virtually all overhangs into doubt. Now it suddenly says fifteen would be fine.

    “BTW, could someone explain to me how the court putting a cap on extra extra seats is more proportional?”

    They set a cap on overhangs, but not on additional seats restoring proportionality. I too found the cap arbitrary and confusing. Did you read the reasoning behind it, rici?


  23. Epsen,

    Thanks for your feedback and for your good description of your proposal.

    A couple of points to which I’d like to respond immediately. It’s a little late for me to give a fullsome response including raising new topics. The following relates more to direct refutatation:

    “Currently, German MMP does not however give an incentive to de-link one Land list from another, since overhangs are all added to the total of each list.”

    This is actually completely untrue. The negative vote weight in fact encourages de-linking. I believe, the CDU would have had a negative vote weight in Bavaria had it not been for the fact that its two component parties are split.

    The fact that party votes have not been de-linked more previously should be a sign that parties are much less likely to pervert an MMP system in Germany, despite the clear benefit.

    This is further supported by German parties’ reluctance to introduce decoy lists to take advantage of the two-vote MMP issue.

    “Compared to yours this is correct, unless parties started strategically de-linking their lists due to the new incentive to do so. ”

    Your comment leaves out on one point. Both systems would still reward strategic de-linking. Your system would still result in a benefit from strategic de-linking. The benefit would only be less in your system, because it’s inherently less proportional to begin with. However, assuming both systems were subject to strategic de-linking, your proposal would still be less proportional in absolute terms (I will work out a good example to provide a response in a day or so, if this is not immediately evident, but my current level of sobriety does not support that).

    “the court deemed negative vote values for parties to be unconstitutional, which seemed to throw the validity of virtually all overhangs into doubt. ”

    Negative vote values do not exist solely due to overhangs. It was strictly due to the fact that the previous German national allocation system did not account for the first past the post seats won by the party.

    For instance, assume everything in the old system had remained constant, except when the national seat total was allocated the starting “n” in the v/(n+1) was equal to the SMDs won. This would in fact eliminate the negative value weight.

    While I acknowledge that your system equally eliminates the negative value weight, it really is only because the initial “n” has been changed from 0 to the number of SMDs won. (Please note that any iterative largest remainder method would have the exact same result).

    Overall, your premise is contingent on the fact that somehow a national allocation system would result in a perverse incentive to de-link, and that this is worth increasing the system’s gallagher index significantly to offset that incentive.

    However, this incentive has always existed, and has only been used by the CSU to date. So there’s no empirical evidence to justify that fear.


  24. Espen, the Constitutional Court says: “Die Wahlrechtsgleichheit und die Chancengleichheit der Parteien sind bei einem Anfall ausgleichsloser Überhangmandate im Umfang von mehr als etwa einer halben Fraktionsstärke verletzt.” which I think means that they think that a deviation from (party) proportionality of 15 seats (halben Fraktionsstärke) is acceptable, but more than that is too much. The Fraktionsstärke is, not coincidentally, the value of the 5% threshold, so my interpretation is that the Court believes that proportionality can be approximated in units of about half a parliamentary fraction. In other words, the 5% threshold is justifiable because any party which fails to meet it would also fail to achieve status, and therefore over-representation of a party (due to overhangs) of about this amount is also acceptable, given the need to balance.

    So the issue is not growth of the Bundestag (there is a comment about this, though, which indicates that they are concerned with the practical aspects of unrestricted growth), but rather proportionality of representation.

    I didn’t read it as requiring a cap, although I suppose that a cap would be one possibility. I don’t see how you could implement that, though: you would have to declare some constituency elections invalid if they exceeded the overhang. In any case, they are looking for a solution which will counter the tendency to an increased number of overhang seats. They consider it likely that the number will continue to increase, so they see the 2009 election as part of a tendency, rather than an exceptional event. So I don’t think they would be favourable to a proposal (like yours) that would only achieve the 15-overhang guideline in the case of the 2009 election but would exceed it in the future.

    I have to say that my understanding of their reasoning leads me to believe that a 2.5% threshold would be much more reasonable than the existing 5% threshold.

    The issue of negative vote weight is separate from the issue of overhangs, and I think your proposal does not solve this problem. So I don’t see how it would be acceptable from that viewpoint either.


  25. Rici,

    The only means I see by which Epsen’s proposition might have a neagtive vote weight would be where votes in one state where a party had an overhang would push the national hare quota higher resulting in that same party possibly losing out on a “hare quota” seat in another state.

    Is this where you’re going with it? I can reason a reason a situation, though it would likely be remote. I guess it would still be unconstitutional.


  26. Nick, precisely.

    The reason this is not as unlikely as you think it is is that in a state where the nominal election results in a lot more seats than the proportional distribution would, quite a few votes can be added without altering the total seat allocation. On the other hand, you probably don’t have to change the pseudo-Hare quota much to reduce the allocation in some other state.

    Of course, this only works with “new” votes; transferring a vote from some other party doesn’t affect the pseudo-Hare quota. Under German law, you cannot abstain from just the list vote. However, consider a CDU voter in a Land with a large number of constituencies in which the CDU is a walk-in, but suppose that the voter lives in a different constituency in that Land; one in which the CDU candidate doesn’t have a chance. Now, I claim, that voter could plausibly make the strategic decision to abstain. Her abstention from the first vote has no effect, since her candidate isn’t going to win anyway. Her abstention from the list vote will have no effect on CDU representation in her Land, which is determined by constituency elections. But it might reduce the pseudo-Hare quota enough to give the CDU more seats elsewhere. (It might help other parties too, but lowering the quota helps parties with a large vote proportionately more.)

    I believe the Court would frown on a system in which strategic abstention was reasonable, since it is an instance of the negative vote problem. But I could certainly be wrong.


  27. Rici,

    Thank you for that summary under no. 77, which is the only one I have read making any sense at all. It still seems terribly arbitrary, but I can see their reasoning. Crucially then, the court did not simply clarify the bounds within which its previous ruling could legally be implemented, which I assumed had created a loophole (that could have facilitated compromise somewhere between the political blocs). Instead it created an additional and separately applicable constitutional standard regarding disproportionality itself.

    If this is correct, of course I agree that as a consequence my proposal is inadequate in its current form. As you both have concluded, it runs afoul of negative vote values and almost certainly of the new disproportionality standard as well.

    However, since your proposal lets the CSU keep its uncompensated overhangs and creates a new incentive for some Land lists to de-link from each other (see my response to Nick below), it too may be in violation of the proportionality standard. If the CDU de-linked only two predictably overhang-prone lists in 2009, in addition to the CSU, this would have given a total of 17 uncompensated overhangs.

    I am coming to the conclusion that we must all either exclude the weakest constituency winners (who instead could be first in line to fill vacancies), or move to some kind of full compensation of overhangs through Ausgleichsmandate (possibly only at the Land level, which would be horrible). One could probably also find ways to chip away at more of the overhangs themselves, or change the system in more fundamental ways by forcibly linking or de-linking lists or by cutting more categories of list votes. Back to the drawing board.

    Politically, it may however in the end be hard to avoid full national compensation. If the CDU/CSU now realises this, the least controversial and most expedient solution is probably to simply graft this on top of the seat allocation system that existed in 2009.


    On de-linking Land lists: We may be talking past each other due to the word “incentive”. Today, it may occur that in retrospect a party realises it could have achieved luckier allocations by de-linking some of its lists. The party will not however be able to predict ahead of the election which particular lists to de-link in order to reap that potential reward.

    In 2009 the CDU actually lost a list seat in Niedersachsen by not being linked to the CSU (the latter would have kept the current seat as an overhang). Likewise, in 1994 the CDU missed out on another net overhang (both under Hare/Niemeyer and Sainte-Laguë) despite Bavarian seats not being affected. In 2002 the CSU would have lost a seat to the CDU in the internal allocation. These effects which by themselves basically amount to rounding errors in either the national or internal allocations can just as easily backfire as be beneficial.

    So I can see negative vote values here, which is important in principle, but I see no incentive to change behaviour.

    This marginal effect is completely overshadowed by the potential reward associated with overhangs themselves, which in addition often can be predicted with accuracy. But today, no list seats in other Länder are cut simply because overhangs occur somewhere else. Under your and rici’s proposal they would be cut, and I see being created a very large incentive to de-link overhang-prone Land lists.

    As you say, the upside of optimally de-linking would roughly be the same in my proposal, but the downside of NOT doing so would be much, much lower (at the admitted expense of overall proportionality). Given that there will be political and mathematical risks to de-linkage, I still think that parties would be less willing to game the system in my scenario. We will never know for certain.

    Under no. 78, you have a very good point on my national Hare quota.

    Regardless, this all feels very unreal now that my proposal has been ruled unconstitutional by a unanimous panel of three at Fruits & Votes!


  28. Alright, so let’s assume we’re working with some modification of the 601 seat version (national allocations).

    How do we address the constitutional court’s limitation on overhangs.

    I think we end having to invalidate constituency elections. This would be particularly true if “vote doubling” becomes a common strategy of CDU voters by voting FDP in the second tier. Or even worse, decoy lists.

    This type of system would reduce MMP to PR but with a more direct election mechanism (choosing a significant portion of members through election).

    The other option is unfettered assembly growth, which would in theory would eventually reqiure a cap in which case, a mechanism for invalidating candidates would be required.

    Let’s say no extra seats are allowed and the assembly size is fixed at 598. This is not a requirement, I’m just using this for argument’s sake. You could instead create a cap of 30 extra seats, with at most 15 being allocated to overhang mandates and 15 being used to balance. But again, you would still require a mechanism to eliminate some SMD winners.

    In 2009, how would we eliminate the 3 overhangs allocated to the CSU in the last election?

    My proposal, eliminate those with the lowest proportion of the vote in their ridings.

    An alternative might be lowest: (proportion of votes)-(winning margin).

    Not sure if this is overly simplistic. Has anyone ever heard of a proposed system where SMDs were eliminated; I am pretty sure it’s never been done in practice.


  29. Reuters reported last week that Germany has passed a new electoral law that “grants smaller parties compensation seats for any overhang seats.”

    According to the article, had the new law been in place in 2009 – when CDU and CSU won 24 overhang mandates, which increased the size of the Bundestag from 598 to 622 – “there would have been an extra 49 compensation seats awarded to the other parties and would have thus raised the number of lawmakers to 671.”

    JURIST has additional links here on Germany’s new electoral law.


  30. While reading Germany’s old electoral law, it seems like Germany actually has a compensatory majority bonus like that found in Malta. It has not ever been put into place because no party has won more than 50% of the second [list] votes since after World War II (the CDU-CSU won more than 50% in 1957, but they’re officially different parties). I would assume this hasn’t changed but I’ve yet to be able to find the text of the new law in English or Spanish.

    “If a party whose Land lists have received more than half of the total number of second votes in the electoral territory of all the Land lists to be taken into account does not receive more than half of the seats available according to the distribution of seats described in subsections (2) and (2a), further seats shall be allocated to the Land lists of that party in the order of the highest number of remaining votes until the Land lists of that party have received one seat more than half of the seats to be allocated in the electoral area. In such an event, the total number of seats (Section 1 subsection (1)) shall increase by the difference.”


  31. Pingback: Why we should stop calling the possible next government of Germany a “grand coalition” | Fruits and Votes

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.