Question on overhang seats

Does anyone “out there” know of a case under a mixed-member proportional (MMP) system in which “overhang” seats were responsible for a manufactured majority? Or determined which potential coalition could form after the election?

I have a vague recollection that there was a German federal election in which one of the coalitions’ ability to form was shaped by overhang seats. I also know that the last Scottish election featured a manufactured majority, but I am not sure if the mixed-member provisions, per se, were the cause, or if the small size of the regions in which compensation is carried out would have been sufficient to produce the outcome.

By “manufactured majority” I mean more than half the seats for a party (or pre-election alliance cooperating in the district races) that had less than half the votes.

By “overhang” I mean… what, exactly? It seems some folks use the term to refer to any case of a party having more seats via the (nominal) district contests than its proportional entitlement would have been, while other folks mean additional seats added to the legislature as a response to such district wins. Either way, the question is of interest.

_____________ UPDATE

The question arose out of a conversation I was having with John Carey, who passes along a much more thoughtful version of the question, which I am posting here:

Can anyone identify a case where a less-than-fully-compensated overhang produced a parliamentary majority in an instance where a fully-compensated overhang (or just a fully proportional election) would not have?

Explanation: The question refers to mixed-member compensatory systems, like those used in Germany, New Zealand, or Bolivia. In instances where a given party wins more seats in the SMD tier than its list PR vote would warrant, the excess seats are referred to as “overhang” seats. Overhangs can be dealt with various ways:

Overhang Expansion: The party with the overhang keeps its excess seats. Other parties are awarded the number of seats each would be entitled to according to its share of the list PR vote. The size of the parliament is increased by the number of overhang seats. (Example, Germany before 2013.)

Full Compensation: The party with the overhang keeps its excess seats. All other parties are also awarded additional seats such that the overall distribution of seats achieves full proportionality. Parliament is thus enlarged both by the overhang seats and by the additional seats given to other parties in order to balance the overhang. (Example: Germany as of the 2013 election, following the Constitutional Court-mandated reform of 2011.)

Non-Expansion: The party with the overhang keeps its excess seats. Seats to which some other party or parties would be entitled, according to their share of the list PR vote, are not awarded to them in order to maintain the size of parliament at a fixed number of seats. (Example: Bolivia)

So, effectively, the question is whether use of either the first (Overhang Expansion) or third (Non-Expansion) method yielded a winner’s bonus that produced a parliamentary majority that would not have prevailed had Full Compensation been used.

32 thoughts on “Question on overhang seats

  1. I’m not really sure if I would describe Scotland’s mixed-member system as mixed-member proportional. I would consider MMP to be a system where list votes are used to create an overall proportional number of seats for each party, and then district seats are subtracted from this figure to produce the number of list seats. Scotland’s system is a bit different. In Scotland, voters do indeed have two votes, but when the list seats are calculated, the list votes for a party are divided by the number of seats already won by that party plus one. There’s a better explanation here: The key difference is that under MMP, the list vote determines the actual composition of parliament, where as under the Scottish system, the list vote only impacts on the list seats. Therefore, the Scottish parliament cannot increase in size, and thus the second type of overhang cannot happen. The first type, though, if it is an overhang (and I personally do not think that it is), certainly caused the SNP to win a majority.


    • It’s not really that there is no overhang, it’s that the overhang is not compensated. I would think that the key to MMP is that the award of list votes is compensatory rather than parallel, and the question then is how to define ‘compensatory’. I think defining compensatory as a system which also compensates overhang to some extent (what you term as “list vote determines the actual composition of parliament, [as opposed to] only the list seats”) is too strict. Scotland, Wales and Greater London are not the only examples where the assembly cannot grow as a result of overhang; Lesotho and Bolivia are others, (and, I suspect, Venezuela’s variant was, too).


  2. In German Federal elections “overhang” seats were not responsible yet for a manufactured majority. Overhang seats always helped the legislative majority to grow even more, though. However, in the 2009 state election in Schleswig-Holstein overhang seats were responsible for a manufactured majority i.e., they have been necessary for the CDU/FDP incumbent government to get reelected.

    For those of you who can read German see here for more details:


  3. In the most recent election for the Scottish Parliament, using 8-region MMP, the Scottish National Party won a majority of seats on a minority of votes. Was this because of using eight regions?

    Not really. Using eight regions gave the SNP one overhang seat. The real problem was that Scotland uses the “highest average” calculation, which Germany abandoned decades ago; it favours large parties. The other methods are “highest remainder” which has the virtue of simplicity, and “Sainte-Lague” which is the best mathematically. A new study has just shown that, if Scotland changed to Sainte-Lague but kept its eight regions, that would generate the fairest result, and the SNP would not have won a manufactured majority.

    Click to access ScottishReport2011_FINAL.pdf

    By the way, even the German federal parliament, for decades, only half-compensated for overhangs, failing to add “balance seats” unlike the state legislatures.


    • From reading the report, it seems either a formula or district magnitude change would have prevented an SNP manufactured majority by several seats. ‘Fair’ is of course subjective. If by ‘fair’ one means a result as proportional as possible, MMP with nationwide allocation (with no threshold) would come closer, whatever the formula.


  4. The procedure used in Scotland certainly qualifies as MMP. The mechanism Henry describes in the first comment does not impact only the list seats, but rather the total number of seats. It ensures that a party that already has seats from the nominal tier has those seats counted towards the proportional entitlement. Ergo, compensatory allocation, a key defining feature of MMP.

    As to Wilf’s point, district magnitude matters far more than allocation formula, although the smaller the magnitude the more the formula matters to final proportionality. In the case of MMP systems, the relevant magnitude is a region in which compensation occurs–i.e. the number of both nominal and list seats in a self-contained allocation area. In either “overhang expansion” or “full compensation” (in John’s helpful terms), that number is not fixed: the magnitude can be increased to increase proportionality, relative to “non-expansion”.

    In assessing the impact of overhang provisions, the key question is what would happen under an otherwise identical pure-PR system, i.e. holding constant the initial magnitude of the regions (or assembly, if nationwide allocation) and the PR formula, as well as the list votes per party.


  5. And for the benefit of those who have already commented or read the initial post, let me note here that the post has been expanded (so to speak) with some clarifying points sent to me by John Carey.


  6. JD is correct that non-expansion was also the rule in Venezuela’s MMP (which has since been replaced by MMM, in which it is indeed true that list votes affect only the list tier, and are not allocated in consideration of the nominal-tier outcome).

    We could add Albania to the non-expansion list. Albania now uses pure PR, but had used alternately MMM and MMP in various elections before 2009.


  7. I recall reading a comment on this blog that it was in the 1994 German federal election that overhang seats came close to having an effect on coalition building. CDU/CSU/FDP won 341 seats, including 12 overhang seats, compared to SPD/Green/PDS which won 331, including 4 overhang seats. If fully compensated, CDU/CSU/FDP would only have 2 more seats than the opposing parties. I think a question then was whether party discipline in Germany was strong enough that the coalition could govern with such slim majority.

    Italy 1996 and Albania 2001 probably qualify as having produced manufactured majorities due to overhang seats, with the usual caveat of there being too few list seats. By the way, as Albania 2005 (and to some extent CDU/FDP in pre-2013 Germany) showed there need not be explicit pre-election agreement between parties that would form post-election coalition in all district races to attempt to gain a majority of seats from overhang…. And then there’s the National-ACT-United Future style agreement in New Zealand 2014.


  8. Italy should not be considered MMP. There was only a partial compensation mechanism, via vote adjustments. As I noted above, Albania used MMP in some elections and MMM in others, and in one of the MMP elections there was the “dummy party” problem, by which a party can effectively turn the system into MMM (same thing that happened in Lesotho once).

    I don’t quite understand jhq’s point about CDU and FDP. They have never cooperated in the nominal tier, to my knowledge. And even in the New Zealand 2014 example, I believe National has candidates in the districts that its partner could (will) win, even though they are there to campaign only for list votes.


    • If a real party uses dummy to win a majority of seats, it counts as a majority with overhang seats, no? Say a party splits into two legally separate parties, A and B. A contest districts only and B list only. So the two parties do have stand down agreement and hence are pre-election allies. Since A receives no list votes all its district seats won are overhang seats. So if A and B combined wins a majority of seats with B having fewer list votes than all other parties over the threshold combined it will be a manufactured majority.

      By the way, I do not have definitive source for the 2001 Albanian election, but a quick look at the results on the Wikipedia page suggests it was not MMM.

      My point about CDU/FDP (and Albania 2005) was that limiting attention to manufactured majority won by a single party or “pre-election alliance cooperating in the district races” with overhang seats may be too restrictive. I imagine there is a spectrum of how two or more parties can act to gain advantage from disproportionality in a not-fully-compensated, two-vote MMP system. On one end of the spectrum, all parties compete wholeheartedly in all the districts and for list votes, and voters genuinely weigh the parties’ and candidates’ credentials in voting (no cooperation). On the other end is that the parties in an alliance don’t compete in the same local district, and, even more efficiently, form a separate super dummy party to gain list votes. Then there are situations in between. Closer to the former end of the spectrum is, say, when a bigger party asking a friendly smaller party (list) voter to give his local vote to its local candidate, with the small party’s tacit approval, because the small party local candidate has no chance to win, with the off chance that the bigger party might win overhang seats. One step closer to latter end of the spectrum would be the two friendly parties doing some wink wink, nod nod at some level, ostensibly for more dignified reasons like to pull a small ally over the threshold, but in reality to win overhang seats, even though they officially compete; and so on. I mentioned CDU/FDP because in 2009, more than a third of FDP (list) voters voted for other parties’ local candidates, and CDU/CSU won a record number of overhang seats. Of course they did not matter in the end, but a scenario where they do was hardly unlikely (until Karlsruhe intervened).

      In the end though, I fully understand if anyone does believe that only explicit agreement between parties counts as “cooperation.”

      I would favor full compensation for MMP, because I’d think it would discourage parties from attempting to milk advantage from disproportionality using whatever tricks in the book; but the possibility of an infinitely sized legislature disturbs me more. Sorry if this post still doesn’t make sense or isn’t relevant. My ADD brain is telling me to erase this draft so I have to click post comment now.


    • National certainly does have candidates in those districts, if for no other reason than there is a electoral advertising spend that is available to electorate (district) candidates. Also, not standing candidates would anger some potential supporters, and likely cost List votes. If the weather is ok at some point in the weekend I might pop ‘next door’ and check out the signs in Epsom.


    • (Since my initial reply is still pending moderation I’ll take the opportunity to rewrite my comment.)

      I looked at some sources on the 2001 and 2005 Albanian elections and it would appear that they were MMP and not MMM, although the district races in 2001 were decided by two-round majority votes rather than plurality. The Socialist Party won 73 out of 100 single-member seats (and no list seats) in a 140-member assembly with 41.4% of party votes in 2001. The Democratic Party won 56 single-member and its pre-election allies won 18 list seats in 2005 with 41.1% of party votes between them. There’s a mention in the blog entry titled “MMP manipulated in Lesotho?.”

      I’d disagree that the use of dummy party turns an MMP system into MMM and disqualifies it as an answer to the original question posed. If anything it is exhibit A for the shortcomings of a not-fully-compensated, two-vote MMP. Of course, one can argue that an MMP system is not really MMP if a district candidate could freely claim he is not linked to a party list even though he may have the party’s label next to his name on the ballot or is in the list itself, something noted to have happened in the OSCE report on the 2005 Albanian election. But surely presenting a party list that contains no district candidate is not impossible to do.

      My point about CDU/FDP was that limiting attention to manufactured majority won by a single party or “pre-election alliance cooperating in the district races” using overhang seats may be too restrictive. It could conceivably include a group of parties that, while officially compete in district races, declare each other as their preferred post-election coalition partners by the campaign, and of which one party is dominant and has by far the most chance of winning district seats. Then by the mathematical reality of an MMP election, to a well-informed voter who wants to see the coalition win splitting his votes is the dominant strategy, whether or not he was explicitly told to do so by the parties.

      In this sense dummy party strategy is but an extreme form of a spectrum of party coordination in MMP. Two legally completely separate parties that form out of one party and both field local and list candidates can disguise the fact that they are employing dummy party strategy by claiming that they are mere allies and their instructions to voters how to vote are perfectly okay strategic voting. Its effect would be only quantitatively different from the scenario of two real parties doing the same thing no so explicitly and to a lesser scale. Any legal recourse that distinguishes them would be arbitrary.

      In an awkward attempt to move my comment back to the original post, I fully concede that my concern about MMP is hypothetical and definitely well studied, which is perhaps why MSS asked for a real world example in the first place.


      • I agree that there is a spectrum here of party strategy. I also agree that the “dummy party” issue is “exhibit A for the shortcomings of a not-fully-compensated, two-vote MMP”. My remark about turning it into MMM was simply that if we have a party that sets up two wings, one for districts and the other for list seats only, in a system without overhang compensation, then its combined seats are the sum of districts won plus a full PR share of the list seats. This is, as I recall, exactly the outcome in the Lesotho case.

        It is interesting that the two cases that experienced this issue took different paths to prevent it in the future. Albania abolished the nominal tier, so it is now pure PR (districted). Lesotho abolished the two-vote feature, but it is still MMP (similar to the first postwar German election). Of course, at the moment, Lesotho has other problems…

        According to the data I have (which started out as Golder’s, but with various additions*), Albania did indeed use MMP in 2001 and 2005, but it had been MMM in 1997. You are also correct, John, about the formula change for the nominal tier: two-round majority in 1997 and 2001 (thus spanning the “MM” era) and plurality in 2005. I was there on an advising mission in 1991, following the first semi-free election (which was just two-round majority); they adopted MMP and used it in one election before the change to MMM (before the change back to MMP…).

        Surely there is a quasi-experimental research design waiting to be done on Albania.

        * some of which come from Carey & Hix, I might add.


  9. Well, I guess that the Scottish system is MMP then! I still think that it is a rather weak form of MMP, and appears to be designed not to rule out the prospect of majority government on a minority of the vote, especially when the opposition to a hegemonic party is split (as in Wales). In Scotland, if NZ-style nationwide MMP allocation was used, in conjunction with a 5% threshold and the D’Hondt method, the SNP would still have won a majority, but a narrower one (SNP 65 (-4) Labour 39 (+2) Conservative 18 (+3) Lib Dem 7 (+2) Greens 0 (-2) Independent 0 (-1)), and there would have been no overhang.
    Personally, I think that the SNP majority was a result of some narrow losses by the Greens, and a good result in the districts.
    As for New Zealand, I reckon that ACT and United Future will get just enough votes to make their electorate seats (if they win them) not overhangs. However, if National+ACT+United Future fall short of a majority, they may have to get the support of the Maori Party, which tends to do well in the Maori electorate seats, but not so well in the party vote. There are two electorates where they are leading (albeit one very narrowly), so John Key’s future may depend on overhangs.
    Also, Matthew, if you have the book ‘The Politics of Electoral Systems’, in the Germany section there is a table of overhangs and their impact on governing majorities. Very interesting.


    • Henry, of course I not only have that book, I have a chapter in it. Plus, I have used it in so many courses that I keep receiving additional copies for free! I did not recall the table about impacts of overhangs, so thank you very much for the reminder.

      And, yes, Scottish MMP is “weak” if being designed to be somewhat disproportional makes an electoral system “weak”; however, in Sartori’s sense, that makes a system “strong”! Either way, to reiterate, the primary reason for its lesser proportionality is not the compensation mechanism per se, but the small magnitude in which compensation occurs (plus the use of D’Hondt, as Wilf noted).

      As for New Zealand 2014, the Maori Party may lose most of its seats.


  10. With overhang seats in Germany and New Zealand, what is the maximum size that the parliaments of those two countries can expand to? Can there ever really be enough overhang seats to fully compensate?

    It is amazing how tight NZ elections are. Governments are won on the slimmest of slim margins. Is there any study on which jurisdiction has the most close elections, and which jurisdictions have the most landslides? It seems as if NZ has the most close elections. It seems as if under MMP in NZ that party discipline has increase more so than under FPTP.

    What about Hungary’s MMM with partial proportionate compensation; 2010 elections and before? There were no overhangs under that system was there? Can there be overhangs under a MMM system?


    • t seems as if under MMP in NZ that party discipline has increase more so than under FPTP.

      Given a Closed List, parties have effective rewards/punishments available to them.


    • No, there can’t be overhangs under MMM, because the very concept of overhang (however they are dealt with under the rules when they occur) is based upon a calculation of a proportional entitlement in the entire compensation area. If there is no compensation, there can’t be a proportional entitlement. Under MMM, by definition, the seat allocation is parallel (adding a given number of list seats onto the nominal-tier seats, which are won independently), not compensatory.

      The Hungarian variant actually is partially compensatory, but the seat allocation is still parallel, which is why it is properly considered MMM (that and its very clear impact on votes-seats distortion). That is, the number of list votes a party wins is based on its list votes, but those votes are adjusted for seats won in the nominal tier. The votes, not the seats (as with MMP), are linked between tiers. Broadly, the Italian system was similar, although the adjustment process differed. (Not an easy thing to explain without examples, so I apologize if that isn’t clear, and refer you to Shugart and Wattenberg, 2001, pp. 13-17, or to the earlier post linked in this paragraph).

      (Various recent changes to the Hungarian system, although the broad outlines I just sketched remain in place.)


    • Rob, only some NZ elections have been close, such as 2005. Most really have not been. Your question on where really close elections have tended to occur is one I have puzzled over a bit, but I was unable to detect any systematic pattern, other than that Cape Verde is a strong contender for the championship.


      • While New Zealand elections don’t tend to be close, most New Zealand governments govern with rather slender majorities. Prime Minister John Key only has 64 seats (with 61 needed for a majority). After 2002, Helen Clark’s government had 62 seats, and in 1999, formal government parties (Labour and the Alliance) held only 59 seats; a deal was done with the newly independent Greens for support on confidence votes. The government coalition formed after the first MMP election in 1996 (National+NZ First) had only 61 seats, the exact number needed for a majority. This coalition splintered, and Prime Minister Bolger was forced into minority government with an ex-Alliance list MP (Alamein Kopu), and five ex-NZ First MPs. I am not sure how this government stayed in office; I assume that there was some deal with ACT or other minor parties.


        • Henry, you are certainly right that the governments have tended to have narrow majorities in New Zealand. That is, of course, a different question from the closeness of the elections. The only two that were close enough in the election result for uncertainty about which party would lead the post-election coalition were 1996 and 2005. The first, as you and Errol noted, resulted in considerable instability and a close call on the final budget before the subsequent election. The second featured some pretty interesting bargaining, but ultimately produced stable and fairly incident-free governance.


    • In New Zealand, the maximum number of MPs for the 2014 election would be 191, if an independent won each of the 71 electorates. 120 seats would be divided amongst those parties above 5% of the vote, and the 71 independents would keep their seats as an overhang.

      In Germany, it’s more difficult to say what the maximum would be, because the mechanism is one that truly earns the “complicated” moniker so often placed on PR systems (see Manuel’s page for a good description:

      As far as I can tell, if independents won all 299 district seats, this system would distribute 598 base seats between those parties above 5% of the national vote. Since none of those parties would have district seats to cause an overhang, the final result would be 598 seats, give or take a few for fractions, distributed amongst the parties, plus 299 independents, for a total of roughly 897 MPs. If this happened, a party would need to win around 75% of the list votes to have a majority in the Bundestag.

      This is worth noting: most systems of accounting for overhang do not seem to take independent district winners (or dummy party district winners who receive few or no list votes) into account. Therefore, it’s possible that, while the distribution of seats between the parties which stood lists is as close as possible to proportional, their share of the assembly would not be proportional to their share of the vote if independents are elected. Fortunately for proportionality’s sake, I don’t believe an independent has been elected to the Bundestag at least since the end of the war.

      It might be possible in some circumstances to make the system cover this by having the independents cover the share of votes below the threshold. So, for instance, in the 2013 German election, 15.8% of the vote was for parties below the threshold. If there were 299 independents, this would mean the Bundestag would have a total of 1892 members, with 41.5% CDU-CSU, 25.7% SPD, 8.6% Linke, 8.4% Green, and 15.8% Independents. If one used the 2009 breakdown for parliamentary parties, the Bundestag would be 6% independent and have a membership of nearly 5,000 members.

      I’m not a fan of this, because this gives people who vote for both a winning district independent and a winning list more weight in parliament than other voters. Perhaps a solution to this would be to take ballots which vote for an independent (or ACT/United Future type party) and to exclude those from the list count if the seat would create an overhang, thus making sure that say, Nationals voters in Epsom don’t get double representation of an ACT electorate MP as well as contributing to electing more National list MPs.


      • Decoy lists are not much of an issue in Germany and NZ. Other jurisdictions that have used MMP, parties have gamed the system so that they can win more seats mutating an MMP system into a MMM. The expansion of NZ parliament toward 191 members seems insane if that were to happen. This is one of the disadvantages of an MMP system. It is not necessarily the best of both worlds.

        Doesn’t the overhang issue different in NZ than say in Germany as NZ uses a nationwide list whereas Germany uses Lander list? How can an overhang be resolve in Germany as each Lander would gain more seats in an overhang situation unless I misunderstand the application of the German MMP system?

        It seems as if having a 50/50 split of MMP or having less list seats than single member districts cause overhangs. Would overhangs be less likely to happen if there were more list seats 55/45 than single member districts?


      • “Perhaps a solution to this would be to take ballots which vote for an independent (or ACT/United Future type party) and to exclude those from the list count if the seat would create an overhang, thus making sure that say, Nationals voters in Epsom don’t get double representation of an ACT electorate MP as well as contributing to electing more National list MPs.”

        I do believe that is the exact procedure prescribed in Germany, although that wouldn’t include ACT in Epsom if used in NZ, since ACT’s candidate is, well, a party candidate, not an independent one.


    • “Doesn’t the overhang issue different in NZ than say in Germany as NZ uses a nationwide list whereas Germany uses Lander list? How can an overhang be resolve in Germany as each Lander would gain more seats in an overhang situation unless I misunderstand the application of the German MMP system?”

      The new system in Germany resolves overhangs pretty well, though it’s quite complex. Here’s a relatively simplified description:

      1. Distribute each Land’s nominal seats (double its number of district seats) by using Sainte-Lague on the second votes within that Land only
      2. If a party won more district seats than the calculation in 1 would entitle it to, it keeps the extra seats as overhangs.
      3. Add together #1 and #2 to get a nominal number of seats for each party.
      4. Divide each party’s national vote total by 0.5 less than its number of seats in #3. This is its quotient. Whichever of these quotients is the lowest becomes the national quotient.
      5. Divide each party’s national vote total by the national quotient, and round the result to the nearest whole number. This is the number of seats it will have in the Bundestag.
      6. Distribute these seats between the party’s Land lists using Saint-Lague.

      This likely means a number far higher than 598 MPs, but it’s a result that’s incredibly proportional to the national vote totals.

      In NZ, the way it works is:
      1. Distribute 120 list seats amongst those parties that won over 5% of the list vote or won an electorate seat, using Saint-Lague.
      2. If a party has won more district seats than the number of list seats the calculation in #1 would entitle it to, it keeps them.

      That’s the end in New Zealand. There is not the quotient process to equalize the number of seats per party, so if a party has an overhang seat, it gets to be overrepresented in Parliament. The same was true of Germany prior to 2013.

      Don’t get hung up on the Laender lists. They basically exist so forming a list is easier than forming a single massive national list. While the initial number of seats allocated to each Land is proportional to its population, the final total does not need to be, as under German law, MPs represent the entire country, not just their Land or district. Therefore, it’s quite possible that Bavaria will be overrepresented as compared to other Laender. However, the nationwide distribution of seats would be unchanged.

      One semi-realistic possibility for a massive German Bundestag would be if the CSU somehow received less than 5% of the national vote yet still won 30 or more district seats. In this case, the overall size of the Bundestag would be massive–if the CSU got 4% of the vote and 30 district seats, it would be around 750 seats, and if they got only 3% of the vote and 30 district seats, it would be around 1000.


  11. I am not sure how this [1998/99 NZ] government stayed in office; I assume that there was some deal with ACT or other minor parties.

    Looks like the May 1999 Budget passed 61 to 59. Ayes National 44, ACT 8, United 1, and 8 independents (or their current party wasn’t elected in 1996). Noes Labour 37, Alliance 12, NZ First 9, Independent 1.


  12. Pingback: On District-Ordered Lists: in reaction to Éric Grenier’s proposal | 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.