Perusing the results of last Sunday’s German election (thank you, Adam Carr), one thing that jumps out at me is the high–by standards of Germany’s proportional system–disproportionality.
The CDU/CSU won 239 seats, out of a total of 622, for 38.4% of the seats. These parties (which we can, following convention, treat as one in national politics) won 33.8% of the party-list vote. So they had an Advantage Ratio of 1.137. That is rather high, given that the proportionality in Germany is calculated on a national basis. It is higher than the CDU/CSU had in 2005 (1.046) or the SPD in 2002 when it was the largest party (1.081). In fact, not since at least 1969 has the largest party ever had an advantage ratio over 1.10, a level that was reached only in 1990 (the first post-unification election, which was held under slightly different rules). So the CDU/CSU did exceptionally well in this election, beating the limits of proportionality by a good margin.
The greater leading-party advantage ratio is not simply attributable to more below-threshold votes. The combined vote for parties that individually failed to win 5% of list votes in this election was 6.0%, which is not exceptionally high. For instance, the total sub-threshold vote was 5.9% in 1998, although it was lower in the two intervening elections (3.0% in 2002, 3.9% in 2005). (In the last pre-unification election, 1987, it was only 1.3%! Very low sub-threshold shares were typical of the long period of stable three- or four-party politics in West Germany.)
Even without the sub-threshold votes included in the denominator, the CDU/CSU Advantage Ratio would be 1.067, which would remain high for a national proportional system.
Of course, a key feature of Germany’s proportional system is its mixed-member (MMP) character, and specifically its provision for “overhangs.” If a party wins, at the nominal tier of single-seat districts, a number of seats greater than its party-list votes would entitle it to, it naturally gets to keep these extra seats. Seats are then added to the parliament in order to compensate the other parties, but even after compensation the party that has won the overhang keeps some of the advantage. I think of it as a reward for local strength, and thus as a natural feature of a system that is as much mixed-member as it is proportional.
Because of the additional compensation feature, the new German Bundestag will have 323 list seats and 299 nominal-tier seats. The total of 622 is eight more than in 2005 (when the starting point was, as this year, 299 in each tier).
Much of the overhang appears to have resulted from strategic coalitional ticket-splitting: some substantial share of voters chose the CDU or CSU local candidate, but gave their party-list vote to the likely coalition partner, the FDP. Of course, this is by no means the first time we have seen such voting, but the scale of it this time seems rather large.
Consider that the FDP obtained 14.6% of the party-list vote, but only 9.4% of the nominal-tier vote. (I know the list vote is its highest ever, by a good margin, and I am sure the nominal vote is also its highest, but I do not have such votes immediately available for all past elections.)
It is worth noting that this sort of strategic ticket splitting would have contributed to the voter’s favored coalition only if the single-seat district in question would be an overhang for the bigger party. It would be interesting to know whether ticket-splitting was systematically higher in these overhang districts (and perhaps in some near misses), as the strategic coalition splitting argument would imply (at least if we assume voters are sufficiently informed to know the shape of the contest in their own district).
So the likely next government of CDU/CSU and FDP will be formed by parties that won a combined 48.4% of the list votes and 48.8% of the nominal votes. Not much difference there. It will have 332 of the 622 seats, or 53.4%. Even if we (inaccurately) treated all of this vote as a joint governmental vote, we would still have an advantage ratio of 1.103. I do not know how that compares historically, but it must still be on the high side.
Did the overhang seats make the difference between a CDU/CSU + FDP majority and a “hung” parliament? I do not think so, but it was a close call. I hope others who have followed the German election can offer their insights on this “disproportional” outcome. (Again, disproportional only in the context of Germany’s highly proportional rules.)
Recall the court ruling that changes must be made to the overhang formula before 2011 (ie: before the next likely election). See Espen’s comment at a previous planting and a recent article in Der Spiegel. How should it be changed?
Espen notes the idea of adding even more seats, so that proportionality can be restored. For example, assume a Land of 100 seats, where the CDU won 25 constituencies on 20% of the vote. We would add enough members so that 25 = 20% * N, giving a new size of 125. But this has two major problems:
1. If some Lander are thus given more seats, the value of a vote will differ between Lander. This is a problem the current system shares, but the proposed change would make it even worse.
2. If a party wins a very high proportion of constituencies, a large number of additional seats would be needed. For example, the current results (German) have the CDU winning 37 of 38 constituencies in Baden-Württemburg on 34.5% of the vote. Evening things up would require 107 total seats, 33 more than B-W would usually get.
Another idea–which Espen says is preferred by the CDU, but Der Spiegel says is preferred by the opposition–is to calculate compensatory seats on a national level, so an overhang in one Land is balanced out by worse constituency results in other Lander. I guess how this would work is that you first count the nationwide total seats each party should get (based on the list results); then subtract the total number of constituencies, to get the number of compensatory seats; and then distribute the compensatory seats between Lander. But this final distribution would be tricky, since you don’t want to give any Land more seats than it deserves.
The best way I can think of to do this to do this is just Sainte-Lague or D’Hondt between each (Land, party) pair, starting each result at the number of constituencies won rather than zero, and skipping any seat allocations that would result in a party or Land receiving more seats than deserved. This does somewhat distort the distribution of seats per party in each Land, but that’s unavoidable without compromising the number of seats per Land. (As a Canadian, I suppose I value not upsetting federalism more than strict proportionality in each region.) When something like this seemed to be proposed for Quebec, I was worried that a small party could end up winning seats in strange regions–but Germany has a reasonably high threshold and a high number of seats per region, so it’s very unlikely there.
A final option is simply to change the balance between list and constituency seats, perhaps moving to 2/3 list seats. If this was done, even an extreme case like B-W would require just one or two overhang seats.
Seed planted by Vasi — 01 October 2009 @ 08:15
Uncompensated overhangs are “a natural feature” of the MMP system? Not really. Most state models add “balance seats” to fully compensate for the overhangs.
Granted, that’s a bit tricky when national seats are distributed among state lists. But Germany don’t much care about geographic balance. Bavaria has seven regions, and in each region the balance seats are added as required in that region. Since the Munich result has more overhangs (CSU support is lowest there), Munich ends up with extra state deputies. Too bad. And at the national level, the compensatory list seats are calculated at the national level, then allocated within the party to the state lists in proportion to their votes, so states with a higher turnout get extra seats. The overhangs then apply at the state level. Example: Saxony-Anhalt, in old East Germany. With ten local MPs in 2004 you might expect ten list MPs. No, they got 13, because the SPD swept all ten local seats, generating three overhangs. But this time, when they had nine local MPs (population loss), they elected only eight list MPs after the Left Party and the CDU split the local seats nicely. Too bad.
Therefore, Germany could live with the same regional bonuses Bavaria lives with, with balance seats added where needed. Why is it only the federal level that has no balance seats? An uncorrected defect in their model — but hardly a natural one.
Seed planted by Wilf Day — 01 October 2009 @ 10:08
Thanks for the comments on the German election. It really does appear that German federal elections are becoming less proportional – before unification, Germany had a very low degree of disproportionality. From what I understand, the increase in Überhangmandate has a lot to do with the ‘new’ eastern federal states (Länder), which tend to be smaller than the western ones, so it’s less likely that a party’s territorial concentration of support will be balanced off by single-member constituency results somewhere else (it’s harder for a party to sweep up seats when there are more of them).
This is seen as a serious problem by some people in Germany, and the courts have ruled that it needs to be addressed. Even though proportional representation is not actually cited in the Basic Law, the courts see recent election results as problematic. One solution, as Wilf Day has pointed out, is to use ‘Ausgleichmandate’ or equalisation seats, as some Länder do, but this will increase the size of the assembly even more, and the Bundestag is already a very large assembly (indeed, it was reduced in size a few years ago, so this might be counterproductive). I’m not up-to-date on recent developments in this matter, but it would be interesting to hear from people reading this who are.
As many people probably know, New Zealand’s MMP system also can generate Überhangmandate, although there are currently only two (from the Maori Party) out of 122 MPs. Our MMP system in Scotland does not allow for any extra seats, and the same goes for the Welsh Assembly and London Assembly.
Tom Lundberg, Glasgow University
Seed planted by Thomas Lundberg — 01 October 2009 @ 10:56
It is pretty clear that unification contributed to both an increase in sub-threshold party list votes, and to overhangs.
Consider the following means of Advantage Ratio (A) for the largest party and total sub-threshold vote percentage.
Last 5 elections before unification (1972-1987): 1.02, 0.8.
Most recent five elections (1994-2009): 1.08, 4.5.
From just these averages, one might assume it was just the sub-threshold votes (and are they higher in former GDR states?) that was contributing to increased A. However, as I noted in the main planting, A is on the rise even independent of sub-threshold shares, and 2009 stands out as especially high in the bonus to the large party.
Again, while I respect arguments that MMP should be designed to be just P, I stand by my (normative) argument that the MM portion of the system implies that parties that are especially good at winning nominal-tier seats “should be” rewarded for it. And that is even before one gets into issues of federalism and potential exacerbation of regional vote-value differences, which some of you have mentioned already in this thread.
I do think that, in New Zealand’s case, overhangs produced by the set-aside nominal-tier seats for the Maori are an anomaly that does not fit the MMP model very well.
Seed planted by MSS — 01 October 2009 @ 12:39
Thanks, Vasi, for luring me out of my lurker-lair!
I will just add that since Ausgleichsmandate could and would be calculated at the national level, there is no reason to limit them to Länder with Überhangmandate. The additional seats within each party could be distributed according to the Sainte-Laguë method (which unlike the pre-2008 Hare-Niemeyer method does not allow for an Alabama-paradox). This would be a simple extension of the present model.
In 2009 this would have given a 666-669 member Bundestag, depending on the rounding method. If one wanted to avoid all these additional seats by instead choosing intraparty malapportionment, a separate solution would have to be found for the CSU because their four Überhangmandate could not have been drawn from other Land lists.
Also, as Der Spiegel points out, the CDU was clearly opposed to changing the law before the 2009 election for obvious reasons, but must now address the issue before 2011 unless they want to tempt a Court-imposed solution. It is worth underlining that the unconstitutionality was votes having a negative value, not votes having unequal values. So it is possible that creative solutions can be found that maintain some disproportionality, but that could in principle remove the system from pure MMP. Within the coalition, the FDP would be guarding the rights of smaller parties so I suppose that is unlikely.
The renaissance of Überhangmandate since reunification has at least four explanations (in addition to tactical voting which I doubt has increased):
- A population decline in the East (without a decrease in constituencies until 2002).
- A somewhat lower turnout in the East.
- A lack of political-geographic diversity within many Länder, i.e. sweeps can occur when the largest party is only slightly ahead of its largest competitor(s).
- A more split vote in the East due to the strong regional presence of the PDS/Linke, leading to many Direktmandate being awarded on low percentages.
All of which lead to seats being won on the cheap without having enough list votes to cover for them. Some of these factors have long been present in areas of the West as well, for instance in Baden-Württemberg (except that this is a growth area, and the strong regional presence is the FDP and the Greens).
In fact, because the SPD fell so far behind even a weakened CDU/CSU and because the smaller parties did so well, the fourth factor was at play in much of the country at this election, leading to so many Überhangmandate.
Seed planted by Espen Bjerke — 01 October 2009 @ 21:09
Tactical ticket splitting works only if a party successfully sweeps all the seats in the districts (or comes close to it). It is rather improbable, even in strongholds of major parties like Bavaria, for a single party to obtain much larger than 50% of the list votes, which would assure winning meaningful numbers of overhang seats. Of course, let’s not forget that this tactical ticket splittting is taking place in context of the more conventional ticket splitting, too…whereby voters are voting for the larger parties in the district and small parties in the list.
Seed planted by Henry Kim — 01 October 2009 @ 21:35
Actually, the Free Democrats had a higher share of the first i.e. constituency vote in 1953 and 1961, and also in 1949, when there was no second vote.
The increase in the vote for parties outside the Bundestag comes from the Pirate Party, which polled two percent of the party list vote – not at all bad for a newcomer. In fact, they been compared with the Greens, who won 1.5% of the list vote in their national electoral debut back in 1980 (the Greens 2.0 anyone?). Quite interestingly, the Pirates had their best showing (six percent) in the constituency of Berlin-Friedrichshain – Kreuzberg – Prenzlauer Berg Ost, where the Greens hold their only direct mandate.
By the way, my website’s Germany page has federal- and state-level results for all Bundestag elections since 1949, as well as aggregate totals for the former West and East Germany since 1990; in fact, I’ve had preliminary 2009 Bundestag election results available since Sunday evening Pacific Time…
(I also wrote a piece for Global Economy Matters on the election, available here.)
Seed planted by Manuel Alvarez-Rivera — 02 October 2009 @ 01:20
It looks like Germany has become Weimarized? Or that the German party system moving toward a Scandinavian style party system.
It looks to me that the Left Party is simply a regional party similar to the CSU in Bavaria except that the Left Party competes with the Social Democrats everywhere else now.
Seed planted by Suaprazzodi — 12 October 2009 @ 18:58