Namibia held its general election for president and assembly on 27 November. Unlike in neighboring Botswana, the longtime ruling party was not voted out of power. Nor was it reduced to less than a majority of seats, as happened in South Africa earlier this year. However, SWAPO did suffer a substantial erosion of its position like many incumbent parties around the world in 2024.
The party won 51 seats out of 96 elected (104 total), which represented a 12 seat drop from 2019. Namibia uses proportional representation (PR) in a single nationwide district with no legal threshold. So seat shares very closely reflect vote shares. The seats SWAPO won are 53.1% of the total elected, and this came on a vote percentage of 53.4%. Yes, slightly under-represented, which can happen with a big district magnitude, no legal threshold, and the use of Hare quota and largest remainders.
A decline in SWAPO’s position is not a new phenomenon. Consider that as recently as 2014 the party won 80% of the votes and 77 of the 96 elected seats. In 2019 it had 65.5% of the vote and 63 seats. So 2024 continues a trend. It now has a cushion of fewer seats over the majority threshold than its losses in each of the last two elections. So if this continues–and obviously it may not–Namibia will have a minority or coalition government for the first time after the 2029 election. Namibia has been independent since 1990, and SWAPO has won every election since it first participated in the constituent assembly election in 1989 (when it won 57.3%).
The presidential contest gave SWAPO a bit more cushion. Its candidate, Netumbo Nandi-Ndaitwah, won 58.1% of the vote and becomes the first woman to lead Namibia. (The presidency is elected by majority runoff, although no runoff has been required thus far.) I would generally expect the president to be elected with a higher vote share than his or her party, especially given PR (and concurrent elections). However, that was not the case in 2019 when the president won only 56.3%, thus running around nine percentage points behind the party. The new president, on the other hand, ran well ahead, as indeed did the president in 2014 (86.7%!).
As of 2024, SWAPO remains dominant, even if its dominance is somewhat more precarious than before. The second largest party won only 20 seats and the third largest just 7. Thirteen parties won at least one seat (8 of them won just one). Namibia remains for now one of the few democracies to have thirty or more years with the same party winning over half the votes at every election.
The French National Assembly is scheduled to vote tomorrow (4 Dec.) on a vote of no confidence against the government of Premier Michel Barnier. The premier has been in office only three months, having been appointed following (with some lag) the snap election President Emmanuel Macron had called in July, and which–predictably, given the timing–did not go so well for the president.
Bariner’s appointment was Macron’s attempt to find someone acceptable to a broad enough range of parties in the assembly such that the government could survive as a minority cabinet by negotiating policy with various other parties in a highly fragmented assembly. In theory it should have worked. Macron is a centrist, and as I said at the time, it was unlikely that the far right and the left would combine to oust Macron’s premier soon after the left had cooperated with Macron’s party against the far-right National Rally (RN). Well, that is precisely what they are doing, unless there is a surprise before the vote: combining on a negative vote to oust the incumbent despite the initiative in appointing a replacement resting again with the president.
The steps that have led the French political system to this point lie in a series of disputes over measures that need to be agreed as part of the budget for 2025. Macron and Barnier made some concessions to the RN, on whose support the cabinet has tacitly depended, but it was not enough. So Barnier resorted to the use of Article 49:3 to force the government’s social security bill into law. As we discussed in 2023, the last time this article was invoked, it is essentially a decree provision, but with a cost: the decree is rescinded if the assembly majority votes to remove the government. Thus the government, in invoking this provision, is forcing the choice as: this bill (and the government) or no government (and no on the bill). It is a powerful tool when the premier is confident there is a majority that would prefer not to reopen the question of who should lead the government and is willing to tolerate the passage of a bill (without having explicitly voted for it). But when a majority of the assembly would prefer a different prime minister–even if not a specific agreed one–it is not likely to turn out well for the government using the measure.
Assuming the government falls, there will be another period of the president and the parties haggling over a new government. The president does not have authority to dissolve the assembly again till July, 2025. As far as the budget is concerned, the current year’s budget can be continued, but that is less than desirable in a time of inflation like the present. A background to all this is that the RN leader, Marine Le Pen, is facing potential disqualification from the next presidential election as a corruption investigation advances. France is entering–or really has been in for a while–a period of considerable political uncertainty.
South Korean President Yoon Suk-yeol, elected in 2022, declared martial law. Given the conditions for which martial law is justified in the constitution (see below) were clearly not in place, this was a power usurpation–an autogolpe. Or an attempted one. It collapsed within about a matter of hours, with the National Assembly voting to annul the decree, which it is explicitly empowered by the constitution to do (also quoted below). This is probably not the last word on the whole matter, as an impeachment seems quite likely now. There were 190 members of the Assembly, including some from Yoon’s party, who voted to rescind the decree. A two thirds majority is required to impeach, and that would be 200. (See update below.)
The constitution has these provisions on martial law:
Article 77
When it is required to cope with a military necessity or to maintain the public safety and order by mobilization of the military forces in time of war, armed conflict or similar national emergency, the President may proclaim martial law as prescribed by law.
Martial law shall be of two types, extraordinary martial law and precautionary martial law.
Under extraordinary martial law, special measures may be taken with respect to the necessity for warrants, freedom of speech, the press, assembly and association, or the powers of the Executive and the Judiciary as prescribed by law.
When the President has proclaimed martial law, he shall notify the National Assembly without delay.
When the National Assembly requests the lifting of martial law with the concurrent vote of a majority of the total members of the National Assembly, the President shall comply.
The step in 77:5 was reached quickly, and after a brief delay the president did indeed comply.
One must wonder what he thought he could accomplish. What a massive miscalculation!
Because it is about to become relevant, here are the constitutional provisions on impeachment:
Article 65
In case the President, the Prime Minister, members of the State Council, heads of Executive Ministries, judges of the Constitution Court, judges, members of the Central Election Management Committee, members of the Board of Audit and Inspection, and other public officials designated by law have violated the Constitution or other laws in the performance of official duties, the National Assembly may pass motions for their impeachment.
A motion for impeachment prescribed in Paragraph (1) may be proposed by one third or more of the total members of the National Assembly, and shall require a concurrent vote of a majority of the total members of the National Assembly for passage: except that, a motion for the impeachment of the President shall be proposed by a majority of the total members of the National Assembly and approved by two thirds or more of the total members of the National Assembly.
Any person against whom a motion for impeachment has been passed shall be suspended from exercising his power until the impeachment has been adjudicated.
A decision on impeachment shall not extend further than removal from public office. However, it shall not exempt the person impeached from civil or criminal liability.
Note that even though there is a post of prime minister, the Korean system is presidential, not semi-presidential. While the National Assembly can “recommend” the prime minister be removed, it does not have authority to force it.
Also of potential relevance right now is Article 84: “The President shall not be charged with a criminal offense during his tenure of office except for insurrection or treason.”
When Yoon was elected in 2022, his vote total was somewhat below 50% and the margin was narrow: 48.6%–47.8%. The current National Assembly was elected in April 2024, and was the subject of an earlier entry on this blog. As I said at the time, the result was a strong opposition majority (176 of 300 seats), as “should be expected from standard electoral cycle effects in presidential systems.” And, “Plurality election of the presidency combined with nonconcurrent elections is a good recipe for frequent opposition majorities–even more when the assembly electoral system is majoritarian in effect even if not clearly in design.” (See the earlier post for the points–and a still-unanswered question–about the design of the electoral system.)
South Koreans have a long record of standing up for democracy, and they did so again today, rallying outside the National Assembly as forces ordered there by the President attempted to shut down the session. South Korea has to be considered one of the most successful of the democracies that emerged anywhere over the last 30–40 years. Today it faced a severe stress test, and while the crisis may not be fully over, the President backed down and seems to be the clear loser here. Most likely he will soon be out of power, as he richly deserves (minimally) for his acts.
When the dust settles, I would like to humbly suggest the Koreans take some free advice I offered in the post on the 2024 election: make the electoral system more proportional and the presidential and assembly elections concurrent. They might also consider making the prime minister genuinely responsible to the assembly majority. Such changes would not guarantee an end to crises of executive–legislative relations, but they could help.
_____
Update. Joel Atkinson, on Bluesky, adds this detail: “Impeachment would require at least 8 members from Pres. Yoon’s own party to vote with the opposition. Then it has to go to the Constitutional Court. But 3 of 9 justices retired, and replacements haven’t been confirmed. The current 6 justices are not enough. A min of 7 are required.”
On 27 October, Japan held a general election for its House of Representatives.1 The result saw the governing Liberal Democratic Party (LDP) and its pre-election alliance partner, Komeito, fall short of a majority. In the previous, 2021, election, LDP had won 259 seats and Komeito 32, for a total of 291 in a 465-seat house (62.6%). In the 2024 election, the LDP won 191 seats and the Komeito 24, for a total of 215 (46.2%). Thus Japan joins the set of countries that have seen substantial losses for incumbent parties this past year.
The LDP+Komeito have not been tossed from power, however, as the runner-up Constitutional Democratic Party (CDP) won only 148 seats and no other party won more than the 38 held by the Japan Innovation Party. There is no viable alternative majority in parliament that could replace them, and opposition parties engaged in minimal coordination in the single-seat districts (SSDs) that make up the backbone of Japan’s mixed-member majoritarian electoral system (MMM). The LDP and Komeito, on the other hand, long have engaged in such cooperation. They do not compete against each other in SSDs, although they present separate party lists. According to a table at the Wikipedia page for the election, Komeito had candidates in only 11 of the SSDs, and LDP in 266. There are 289 such districts.2
Komeito won 4 SSDs and LDP 132. Their combined constituency votes amounted to 39.8%. Given that they do not compete against each other in these districts, we should sum their votes at constituency level. In any event, because Komeito runs in so few, its total nominal (constituency-candidate) vote amounted to only 1.35%. In the lists, where they run separately, Komeito won 10.9% of votes and the LDP a mere 26.7%. Notice that the sum of the alliance partners’ party-list votes reaches 37.6%, so only a small degree below the nominal vote. By seats, LDP won 59 (of 176) from party lists while Komeito won 20 (33.5% and 11.4%, respectively). The list tier is not nationwide, but is divided into 11 regional multi-seat districts, which thus have an average district magnitude of 16. Because this is purely MMM, a party’s total seats are a simple sum of single-seat districts won and list seats won in the separate components of the system (“parallel” allocation).
The LDP’s performance is quite poor but not shockingly so from an electoral systems analysis point of view. For instance, if we knew nothing other than the existence of the basic tier of 289 single-seat districts, we would expect from the seat product model that the largest party would win around 49% of them, which would be 142. So they came about two percentage points of the seats short of a pure electoral-systems expectation in that component of the system (including, as noted, the Komeito seats in their total).
If the list component were a stand-alone system, its seat product would be 2816 (that is, 176 seats, times the mean magnitude of 16). For that, we would expect a largest party seat size of 37.1%, or about 65 seats. The LDP came a little short of this, with its 59 (33.5%).3 So really the LDP+Komeito did not do badly at all! Finally, Japan had a normal election for its electoral system!
Yes, we call this system mixed-member majoritarian, but that does not mean an actual majority of seats should be the general expectation. It is “majoritarian” in the sense that it prioritizes winning single-seat districts and thus creates a substantial bonus from votes to seats, relative to the main alternative model among mixed-member systems, MMP (as in Germany or New Zealand). And however you slice it, the LDP won a substantial bonus: The LDP alone has 41% of total seats on 26.7% of party-list votes. It looks less like a bonus based on nominal votes (where it won 38.5%), thanks to the coordination with Komeito in these seats. It should be kept in mind that the very existence of an alliance coordinating SSD nominations is itself a practice made optimal by the majoritarian nature of the system–allowing parties to keep their separate identities in the list component without “spoiling” seats in the district component.4 The combined LDP+Komeito seats at 46.2% also represent a substantial bonus, coming on 37.7% of list votes for the two parties together or 39.8% of nominal votes.
I will conclude briefly by noting that for MMM, calculating expectations the way I did here (treating it as two distinct electoral systems, each with its own seat product) does not normally work so well. MMM is in fact a more complex system than this implies.5 But as an exercise here, it demonstrates well how this result was not especially extraordinary in an electoral-systems sense, even if it was politically unusual for Japan’s experience with MMM to date.
As for government formation, the LDP leader and incumbent PM Shigeru Ishiba won a parliamentary vote to remain in office, 221 votes to 160 over the CDP leader. It will be a minority coalition (LDP+Komeito) rather than a majority coalition. “The Japanese people expressed their strong desire for the LDP to do some reflection and become a party that acts in line with the people’s will,” Ishiba said. With all due caveats about inferring collective desires (let alone “will”) from an election result, that seems like a fairly accurate conclusion. The government plans to negotiate bill by bill for support from other parties.
__________
I am almost a month behind on an election that I would normally have paid close attention to. Too many elections this year–not that that’s a bad thing! ↩︎
That leaves 12 in which neither had a candidate. I note that there were 12 independents elected. I do not know if these are perfectly overlapping sets or not. ↩︎
If we added the Komeito partner’s seats to this (even though we really should not, given they run separately for these seats), the combine would be up to 79, which would be a little above expectation, at 44.9%. ↩︎
By contrast, if it were a more typical “majoritarian” system like FPTP, parties have to choose–either maintain their identity by running against each other in SSDs, or effectively merge, even if only tactically. In most FPTP systems, the former practice is the norm. ↩︎
Work that I am doing but which is not complete yet suggests that the “effective seat product” of this system is around 580. Note that this is quite far from a simple arithmetic average (or even a geometric average) of the two components, treated as having their own seat products (2816 and 289). So the exposition in this post is just an exercise for this election, and not a “recipe” for how to analyze MMM generally. That will come in due time! If 580 is (effectively) “correct” the average largest party size should be 45%, or 210 seats out of 465. That is 19 seats short of what the LDP alone got in this election, and of course the seat product model can’t “know” that the LDP and Komeito act like one party in some respects (but like two in others). And in most Japanese elections under MMM, the leading party/alliance has done much better than this. The bigger point is the LDP’s typical overperformance is due to political factors and not just the electoral system’s “majoritarianism,” per se. ↩︎
Based on preliminary results, the mayor of San Francisco, London Breed, was defeated. San Francisco uses the alternative vote/instant runoff, i.e., ranked-choice voting (RCV) in a single-winner contest. This was the first election in the city to be concurrent with a presidential general election, following the passage of a city ballot measure in 2022 to make such a change. The previous election was in 2019. The following screenshot from the SF elections site summarizes just how fragmented this year’s contest was.
The incumbent managed to get not even a quarter of the first-preference votes. In addition, there were a lot of candidates, including four with over 15% of the (first-preference) votes, two more above 2%, and nine candidates with less than 1% apiece. It took fourteen rounds of counting to reach 50%+1 of (non-exhausted) ballots, but in the end the candidate with the plurality of first-preference votes emerged as the winner. The full table of the fourteen counts shows 54,245 ballots exhausted (not continuing due to not having a preference marked for any remaining candidate) between the first and final counts. That is just short of 15% of the total first-preference ballots cast. This means that the newly elected mayor has just over 47% of the total number of ballots that registered a first preference.1
I do not know if this result shows RCV in a “good” or “bad” light, so I will just leave it there without further comment.
Finally, let’s look briefly at the other RCV contests on this ballot for either citywide contests or districts within the city. Six of the Board of Supervisors districts had contests. In two of them the candidate who was second in first-preference votes consolidated a majority on transfers and won. In those races the split in first-preference votes was 40.8%–39.8% and 31.0%–29.5% (with the third candidate at 19.6%). The other four had initial top two at 46.9-41.7, 39.8–29.6, 46.7-36.5, and 42.1–28.1 (with third on 20.2). So most of them were not very close, but the two closest and one of the two with a leader under 40% went to the initial runner-up after transfers. The other contests were not close at all: the City Attorney winner had 83% of first preferences, District Attorney had 66%, Sheriff 80%, and Treasurer was unopposed.
________
Based on numbers in the linked table at time of posting, 172,082 votes out of 365,122. ↩︎
The past week saw the heads of government in both Israel and Germany dismissing a senior member of their coalition cabinet. German Chancellor Olaf Scholz fired the Finance Minister, Christian Lindner, while Israeli Prime Minister Benjamin Netanyahu fired Defense Minister Yoav Gallant. Both Scholz and Netanyahu head coalition governments. The result of the dismissal in Germany precipitated the collapse of the coalition and will likely mean new elections in the coming months. The result of the Israeli dismissal is probably a strengthening of the position of the current coalition. Why such difference?
The answer is quite simple. But even simple answers can be enlightening. The core of the difference is that Lindner is the leader of one of Scholz’s coalition partners, while Gallant is a member of Netanyahu’s party. By virtue of being head of government, a prime minister (and the German Chancellor is, in generic terms, a prime minister) can be expected generally to have authority to dismiss other ministers. He would seemingly not be much of a head of government if he had no personnel control over the government itself.1 Moreover, a party leader also certainly has authority to shift personnel within his own party.
In Netanyahu’s case, these authorities overlap–he is changing government personnel in a ministry that the existing coalition agreement assigns to his own party. No big deal, actually. No big deal in a legal or constitutional sense.2 It is rather extraordinary to see this dismissal in the midst of a war (and one with remarkable successes recently; it is not as if he is being dismissed because his leadership of military defense is going badly). In other words, in this week’s Israeli case of a dismissed senior minister, it is an intraparty dismissal.
In Scholz’s case, he has dismissed the leader of a coalition partner in the three-party “Traffic Light” coalition formed between his Social Democratic Party (SPD), the Greens, and Lindner’s Free Democrats (FDP) after the 2021 election. This makes it an interparty dismissal. This is an entirely different matter. Coalition agreements rarely, if ever, give one party the right to determine which personnel of a different party to the agreement will fill specific posts. The agreement typically sets which ministerial posts will be held by each party, but leaves it to the party to determine who gets the post. Of course, at formation time, it would be common knowledge who will get the highest ranking post that a partner is agreed to take. The point here is this is no mere personnel matter. This move reflects a deep policy dispute, and obviously one that Scholz had determined was not reconcilable. The coalition is over, as the FDP (as fully expected by all, including Scholz) withdrew its ministers from the cabinet.3 Scholz has appointed Jörg Kukies of the SPD (but not a member of the Bundestag) as the new finance minister.
Germany at this moment has a minority coalition government of the SPD and Greens, with the FDP having joined the opposition. Germany does not have much experience with minority governments, although of course many other parliamentary systems have them frequently. Past minority governments in postwar Germany lasted only weeks. There was one in 1966, headed by the Christian Democratic Union (CDU) which ended when the CDU replaced its leader, who negotiated the first “grand coalition” (with the SPD). The second one was in 1982 when the FDP flipped from supporting a cabinet led by the SPD to one led by the CDU, via a constructive vote of no confidence.4
Germany is certain to go to early elections,5 although not necessarily immediately. Scholz initially proposed waiting till January to hold a vote of confidence, which he would lose, but thereby holding off setting the formal timeline for a new election. Under that plan the election likely would have been in March, 2025. In the absence of a government with confidence of the Bundestag, the (basically ceremonial) president can dissolve parliament, which starts a 60-day timeline to an early election, implying there could be a path to holding an election much sooner. The latest idea is an agreement with the CDU to set the election for 23 February (DW article on 12 Nov., at first link in this paragraph).
The political logic to the FDP’s move6 to the opposition is that the party is apparently desperate to recover its voting constituency. Since entering the current government, its polling has been terrible and it has been above the 5% party-list threshold in few polls over the past year. It also has performed poorly in recent state elections. Even if being in opposition allows it recover in time for an election, coalition formation after the next election is going to be difficult. Currently the Christian Democratic parties (CDU/CSU) are leading, but with just under a third of the vote. The SPD has dropped to third place on 16% (just behind the far-right AfD), and the Greens are down to 10%.7
In the Israeli case, as noted, the dismissed defense minister is a member of the prime minister’s own party, Likud. And he has been replaced by another Likud politician who was already in the cabinet, Israel Katz, who had been foreign minister. In this sense, it is an intraparty change. However, there is also an interparty aspect. The new foreign minister is Gideon Sa’ar, leader of the New Hope Party, who had joined the government in late September but had no policy portfolio till now. Changing which Likud politician is in the defense ministry clears a path for Sa’ar to receive a prominent portfolio.
Gallant, in various ways, had been leading almost a de facto opposition within the cabinet and party,8 so his dismissal was a long time coming. That is not to deny that it was shocking in its timing. But politically, it was probably inevitable at some point that the Likud party leader would impose discipline on him.9 Gallant’s departure buttresses the PM’s own authority, while also giving a senior cabinet post to another party, thus cementing what is now an oversized cabinet (six parties holding 68 of 120 seats) instead of a bare-majority one (64 seats previously).10 New Hope, which had run as part of Benny Gantz’s National Unity list in the 2022 election, currently holds 4 seats.11
These two cases of a head of government firing a top minister are superficially similar, but differ greatly in their likely impact. Much of the difference, as reviewed here, is in whether it the firing was an interparty matter as in Germany or a (mostly) intraparty one as in Israel.
Also, keep in mind that the next German election, which now will be held early, will take place under a new electoral system.
_____________
(Edited these two sentences, as I am told there are actually exceptions to this seeming rule!) ↩︎
This did not stop petitions from being filed with Israel’s Supreme Court claiming it was unconstitutional. The Court predictably swatted the petitions away. Indeed, the Attorney General, who is often at loggerheads with Netanyahu, said in her response to the petitions what should have been obvious: The PM has this authority. ↩︎
Actually, one remains, but how he is doing so only buttresses the point that the firing of Lindner meant the coalition was over. Transport Minister Volker Wissing is retaining his cabinet seat, but resigning from the FDP. The decision to enter or exit a parliamentary coalition is a party decision, and an individual member can dissent from it only by leaving the party. ↩︎
Under a constructive vote of no confidence, a parliamentary majority can remove the incumbent cabinet via a no-confidence vote only by simultaneously electing its replacement. (This provision does not apply when the cabinet itself has asked for a vote of confidence and lost–even if it is an “engineered” loss like the one in 2005 that precipitated an early election.) Incidentally, Israel also has a constructive-vote provision. ↩︎
Prior to last week’s event, the expected next election was going to be 28 September, 2025. ↩︎
Obviously, the specific “move” was Scholz’s, but this happened because Lindner was resisting certain policy priorities of the other parties in the coalition with respect to the upcoming budget. See “confrontation paper” in Stern (in German), for example. ↩︎
Best guess is the parties just named (not including the AfD, of course) will be the next coalition. But it won’t be an easy one to bargain out. ↩︎
On a wide range of issues. He and Netanyahu have had well publicized disagreements over various decisions in Gallant’s defense portfolio (for instance speaking out over the alleged lack of postwar planning and other more recent disagreements), Gallant was influential in stopping (at least temporarily) the judicial reform push in 2023, and this year has been demanding that a bill codifying Haredi exemptions from military service not be advanced without agreement of parties outside the coalition. (On this latter point, Sa’ar seems to agree, as do some other members of Likud, so the firing may not change the political dynamic.) ↩︎
Around the same time various disciplinary measures have been imposed on other Likud members of the Knesset, Dan Illouz and Yuli Edelstein, over their recalcitrance on specific government bills. ↩︎
An oversized cabinet is one that contains at least one party whose departure would not deprive the coalition of its parliamentary majority. A bare-majority cabinet is one in which all parties are essential to maintaining the majority, as was the case with the cabinet formed after the 2022 election (not counting the period under which there was a larger war cabinet in place). ↩︎
New Hope formally left National Unity in March at time when he and Gantz were still members of the temporary war cabinet. Sa’ar left the emergency cabinet in late March. ↩︎
Botswana held its 13th general election on 30 October, and the result was a big defeat for the Botswana Democratic Party (BDP), which had ruled the country through a series of regular elections since independence in 1966. The BDP conceded defeat swiftly.
Botswana uses a parliamentary form of government with a single-seat plurality (or first-past-the-post/FPTP) electoral system. While the head of government is called “president” and serves also as head of state, the system is parliamentary. There is no separate election for the chief executive, who is simply the head of the party that wins the majority of seats (or if none did, the leader elected by parliament). And a parliamentary majority can vote no confidence in the head of government/state (see Art. 92 of the Constitution of Botswana). There are 69 seats, so a majority is 35; however, 8 of these are appointed or ex-oficio seats. Thus the election takes place in 61 single-seat districts.
The new ruling party will be the Botswana National Front (BNF), which headed a pre-election coalition known as the Umbrella for Democratic Change (UDC). The UDC consists of three parties and some independents who coordinated nominations, thereby not contesting districts against one another. It collectively won 36 seats, 23 of which are BNF. (See compilation of district results, although unfortunately it does not indicate which party contested for the UDC in each district.)
Remarkably, the national results show the long-ruling BDP falling to third place in seats, winning only 4. Those four out of the 61 elected seats means it was punished badly by its geographic vote distribution, winning only 6.6% despite still being second in nationwide votes with 30.5%. The parties allied in the UDC combined for 37.2% of the votes and their 36 seats, while a bare majority overall, is 59% of the elected seats. The second place finisher in seats, the Botswana Congress Party, won 15 (24.6%) on 21% of the vote.
That is a whopping 1.59 advantage ratio for the winner (percent seats divided by percent votes), as well as a reversal of the order of second and third in seats relative to their votes. Single-seat plurality can be like that.
I have tended not to include Botswana in my comparative elections work. Many datasets on elections do not include countries in which no alternation in government has occurred, even if the usual democracy-rating organizations say elections in the country are free and fair (as they have for Botswana). It is somewhat difficult to do meaningful analysis of votes–seats relationships when one party is so dominant for purely political reasons (whatever those might be in Botswana). And fifty eight years of rule by one party is nothing if it is not dominance. But now there’s no excuse: Botswana has had a regular alternation via elections.
The expected largest party size for an assembly of 61 seats, elected in single-seat districts is 59.8% of seats, per the seat product model.1 So this election basically nails the prediction. The effective number of seat-winning parties in this election is 2.38, which is downright fragmented relative to the expectation from a seat product this small, which would be 1.98. (The seat product model implicitly “thinks” the second party should win the remaining 25 seats; actually five parties won seats, counting the UDC as one and also counting the one independent not in the UDC as a party.2)
Botswana has almost two and half million people, which is hardly a large country. However, its assembly is quite small for that population. By the cube root law, we should expect an assembly more than double, with a precise prediction of 134. If the assembly were this size instead of only 61 elected seats, the largest party would be expected to win around 55% of the seats, or about 74 of 134. Such a larger size would not have guaranteed alternation or even a minority government in past elections, but surely would have reduced the BDP’s dominance. For instance, in the previous election of 2019, the BDP won two thirds of the elected seats on around 53% of the vote. Such massive seat bonuses as seen in both of these elections would be considerably less likely in a larger assembly. Maybe the “umbrella” could provide better protection for democracy by considering increasing assembly size, although I shall not hold my breath that this would be one of their priorities. They could also consider some form of proportional representation, and apparently doing so actually is in the UDC’s manifesto, although I do not have details.
____________
The calculation yields 36.4 of 61 seats (assembly size raised to the exponent, –0.125), but since in reality it has to be a whole number, we can say the on-average prediction is 36, or just what the UDC won in this election. It is worth adding that while the seats won are as expected for the leading party in an assembly of this size, it won these from a much lower voting base than expected. We might expect such a party to have won around 51% of the votes. The votes were rather remarkably uncoordinated for an alternation-inducing election!↩︎
Actually, under the seat product model, we expect the number of parties of any size–not the effective number–to be 2.8 on average. In other words, at times a small third party would win a seat or two, and at other times just the two main parties would have seats. If only two parties won seats, with 36 and 25, we get an effective number of 1.937. To get closer to the expected 1.98, you might have three parties at 36-24-1, which would be 1.987. (The last of these might be an independent, and indeed as noted one independent who was not coordinating with the UDC did win.) ↩︎
This post is mainly here in case readers want to talk about the election. I am not really sure–as of mid-afternoon in California on 5 November–that I want to. But since I should put something in this space, here we go…
It seems that US Election Day is upon us. I approach this day and the night (and possibly days of uncertainty) ahead with trepidation. I very much fear what the future of democracy will be–and not only in this country–if Trump wins. And yet if the polling is correct in its aggregate, he very plausibly is winning. The polls will not be correct, however. Some degree of error is inevitable. Will the error favor Harris? I guess I think so, in this year’s electoral context. However, I know there is no good objective way to predict error and its direction, or else it would not be an error!
I will not offer a prediction. Just a hope. I want Harris to sweep all of the swing states, or at least enough to win, preferably with more than the bare 270 that this fairly plausible map would produce.
That would be too close for comfort! After all, she could lose, for example, either Pennsylvania or Michigan by a small number of votes. On the other hand, I really do not think her winning both Georgia and North Carolina is a stretch. So the above is NOT a prediction.
In the previous planting, I reviewed the various measures on state ballots proposing to adopt nonpartisan first rounds and RCV runoffs. This is, of course, the model that Alaska has used in recent years (and that is subject to a repeal attempt on this year’s ballot). Sightline has an interesting article in which they note that several candidates who qualified in the “top four” qualifying rounds in various races have dropped out prior to the runoff. This is perhaps a disappointment for advocates of these systems, who assume RCV is spoiler-proof so why would anyone drop out? Of course, a tendency of trailing qualifiers to withdraw is not (or should not be) surprising to a political scientist. Parties adapt, and can be expected to encourage weaker candidates to clear a path for their stronger candidate. This will not be successful in all cases–if there are real intraparty factions, the qualified candidates may insist on remaining in and continue their “primary” fight in the runoff. But they also know something advocates gloss over: RCV is not in fact spoiler-proof. That depends on the order of finish in first-preference votes and the tendency of voters to give full rankings. For instance, Sarah Palin may have spoiled the US House special election in 2022 won by Democrat Mary Peltola when a more mainstream Republican (Nick Begich) was eliminated in the second of three rounds of counting. (Peltola had just under 40% of first-preference votes whereas the two Republicans had 31% and 27%.) Nonetheless, the Sightline article notes that elections are becoming more competitive under the new system, although the average number of candidates in state legislative elections has actually risen only from around or a bit under 2.0 to the range of 2.0 to 2.5. That is probably less of an increase than advocates expected.
Several US states (and also localities) are voting on 5 November on whether to adopt ranked-choice voting (RCV) systems. Most of these are the “unified primary” model, often misleadingly called “open primaries.” I have explained before why I do not like these reforms, as well as why they are actually abolitions of primaries not open (or even unified) primary systems. (I also sketched out in that earlier planting some tweaks that could make me somewhat more favorable to these proposals.)
The short version of my view on these is that they undermine parties and thus make more difficult the voter’s task of identifying the “correct” candidate and figuring out who is collectively responsible for governance (a necessary condition for retrospective accountability). They have these flaws due to being “top four” or “top five” whereby that number of candidates qualify out of a potentially crowded field, regardless of party label. Further, the implication is that the second round will be at least partly an intra-party contest for the single seat, which only exacerbates both of these flaws. This is also the sense that the first round is not a primary at all–as generally understood, a “primary” (including an “open primary”1) is a party’s nomination contest. Thus a first round that can advance multiple candidates of the same party is not a primary by the standard definition. It is just that–a first round of a two-round process. By extension, these measures also do not exactly adopt a “general election” via RCV, but a second round (a runoff) using RCV in what is effectively a two-round general-election process.
A helpful list provided by FairVote indicates that the following states have proposals to adopt this combination of a nonpartisan first winnowing round followed by an RCV runoff: Colorado, Idaho, and Nevada (also potentially Arizona, explained below). This is the model currently in place in Alaska, although in that state there is also a measure in this election that would repeal the system. In Colorado and Idaho, the first round would advance four candidates to the runoff (as is the case in Alaska), whereas in Nevada’s proposal five would qualify (per Ballotpedia).
Oregon also has a measure on the ballot to adopt RCV, but this one is different. Measure 117 in that state would adopt RCV in primary and general elections. Rather than abolishing partisan primaries and replacing them with what is essentially single non-transferable vote (SNTV) where you vote for only one candidate but four (or five) of them advance, Measure 117 would provide that each party use RCV to choose its nominee. Then RCV also would be used among the nominated candidates in the general election. This is the model currently used in Maine. I find it more beneficial, or at least non-harmful.2 I do not know if I would favor it were I voting in Oregon, whereas I am fairly certain I would vote no if I were voting on one of these measures in Colorado, Idaho, or Nevada.
Too often, commentary on these proposals (unsurprisingly including FairVote but also even Ballotpedia) do not make clear how different these approaches are. A nonpartisan SNTV winnowing round is very different from an RCV partisan primary, and an RCV contest that is simultaneously interparty and intraparty is very different from RCV among each party’s single nominee. (See Manuel’s comparative analysis at this blog of the Alaska and Maine elections of 2022.)
In Arizona, there are two conflicting measures on the ballot concerning primary elections (see the official ballot pamphlet). Proposition 133 would give existing partisan primaries explicit constitutional status, and would supersede any local measures that are inconsistent with this requirement.3 Proposition 140 would institute a nonpartisan first round of uncertain format. The second round would require a majority (RCV when more than two candidates). It would require the legislature to determine how many candidates would qualify in the first round. It says only that the law will provide that not fewer than two and not more than five will be advanced, regardless of party affiliation, if the final election is for a single candidate.4 Once enacted, this number could be changed by law “not more than once every six years.” Wow, this is kind of a mess! I guess I’d vote no on both.5 (For fun, I note that the supporters of 133 state that the other measure, 140, is an idea “imported from California.” See the pamphlet, at p. 37.6)
There is also a measure in the District of Columbia (Initiative 83) that would adopt RCV and change primaries to the extent of allowing voters who are unaffiliated with a political party to vote in a party primary.7 Given the dominance of one party in DC, I am not sure what the point of RCV in the general would be, although I can see the motivation to let unaffiliated voters participate in the dominant party’s primary. (Whether that is a “good” idea or not is a separate matter.8) If understand correctly, RCV would also be required in the party primaries.
There are also some local RCV measures. I did not look at them in detail, but they are listed at the FairVote article. It is not clear from the article if any of these are multi-seat STV or if all are “instant runoff” in single-seat contests.
_________
As generally understood at least before the nonpartisan-first-round reformers appropriated the term, an open primary is “open” only in the sense that you do not need to be registered in advance with the party to participate in its nominating election. But the contests in which you will be voting are still about nominating candidates for a specific party. A “closed” primary is one you may vote in only by having registered with that party. ↩︎
The benefits of RCV in the general election are scant if viable parties beyond the top two (or independents) do not exist or emerge. While such parties could emerge given RCV, there is no necessary reason to predict it to happen. ↩︎
Are there partisan local offices in Arizona? It seems this is mainly an effort to preempt any pending or future local RCV measures. In fact, in the ballot pamphlet (p. 40) it is pretty explicit in that the supporters say 113 “would proactively protect Arizona from the horrors of Ranked Choice Voting (RCV).” ↩︎
Arizona’s state legislature includes two-seat districts. For these the number of candidates passed through to the runoff would be set by law to be in the range of four to seven. ↩︎
Or perhaps abstain on 133 and vote no on 140 (given that in event of both passing, the one with the higher total takes effect). ↩︎
I will also add that finding this took me some time. The state’s election website is not well organized! And the pamphlet itself is a jumble with multiple statements for and against each measure that go on for pages and pages. I am used to the very easy to navigate website and brief consolidated arguments in the pamphlet from sponsors and opponents that we get in California. Perhaps Arizona might consider importing this model. ↩︎
The measure calls it a “semi-open primary.” But as best I can tell it is an actual open primary (as defined in the first footnote to this post). I would understand “semi-open” to mean that the party decides whether to allow unaffiliated voters to participate in its primary or to close it to only party registrants. This is the case in California (for presidential nominating primaries–Republicans, for example, run a closed primary and Democrats an open one). ↩︎
I kind of feel that should be a party’s choice, not one imposed on it. But I don’t have a strong position on this question. ↩︎
Georgia (the country on the Black Sea) holds its general election on 26 October. It will be the first election since an electoral reform was passed abolishing the mixed-member majoritarian (MMM) system and replacing it with a fully list system. The electoral reform was passed in 2017, but subject to a long implementation. The reform was part of a package of constitutional amendments that also abolished direct elections for the presidency. The direct presidential election of 2018 (for a six-year term) is thus scheduled to be Georgia’s last.
Thus with this election, Georgia will complete the move to a PR-parliamentary system. Its democratic period since the fall of the Soviet Union began with a president-parliamentary system, was changed to premier-presidential, and now complete the journey to a fully parliamentary-dependent government. The electoral system has been MMM since 1990, although with various changes to specific provisions. The article linked above (first link) about the reforms passed in 2017 implies that the MMM system was to remain unchanged for the 2020 election and then be PR for 2024. However, there actually was a substantial reform–enacted later–to the MMM system that took effect in 2020. In that election there were 120 of the 150 seats elected by party list. In previous elections it was more balanced–77 by list and 73 by single-seat districts (two round majority, which was also the case in 2020). In fact, with a list component comprising 80% of seats, this was already about as “PR” as any MMM system ever has been while nonetheless still being MMM.1 Now they will complete the journey to PR.
However, it should be noted that the full transition to PR actually will be implemented till 2028. The system for 2024 says that seats that would have gone to smaller parties had they not fallen below the threshold shall be transferred to the largest party, so it is actually a bonus system not fully PR. For the 2028 election this bonus clause is scheduled to be removed.2 [Update: It appears that the bonus that had been intended to be in effect in 2024 was repealed prior to the election. Indeed, the item just linked in the preceding sentence actually says that, but I must have confused it with an earlier article about the bonus provision. Moreover, the results show an almost proportional result, notwithstanding two parties combining for over 5% of the vote below the threshold, along with several other smaller parties.]
The new PR system will have a 5% threshold. The threshold on the list component in 2020 was only 1%. (In 2016 it had been 5% for the smaller 77-seat list component.)
As for the head of state, it will still be a president, but one elected through an electoral college. This electoral college consists of all the members of parliament along with local and regional government representatives. Thus there is no sense in which this electoral college retains any element of popular election (in contrast to the US electoral college under longstanding practice). The total size of the body will be 300, so parliament will comprise exactly half the electoral college. The presidential term is being reduced to five years. (This information is, again, at the first link.)
We thus will have one fewer mixed-member system and one fewer semi-presidential system in the world of democracies as of this coming election.
______________
Was the list portion so big that we should consider it a proportional system already in 2020 and not mixed-member majoritarian? Hardly. The Georgian Dream party, which had about 48% of the votes both in both the aggregate of the districts and for the national party lists nonetheless won all 30 district seats (13 in the first round an 17 in runoffs). It won 60 of the 120 list seats–exactly half, for a very modest boost relative to its vote share. This means overall it won 90 of the 150 seats (60%), a substantial majority that was fully dependent on the mixed-member nature of the system and surely “majoritarian” in any sense of the term. (There was a provision in the 2020 law that said a party could not win a majority of seats without having cleared 40% of the votes. So there was a cap on overrepresentation, somewhat reminiscent of the Mexican system. I consider Mexico to be MMM, and this is no less so. In any case, the provision obviously was not trigged.) ↩︎
The linked item phrases the bonus as follows: “the transfer of undistributed mandates (proportional to the votes garnered by parties which fail to clear the 5% threshold) to the winning party.” I do not fully understand how this will work in practice, but it obviously benefits the leading party. ↩︎
On the Wikipedia page for Politics of Lebanon, we find the following information. After the quoted passage, I will explain what is wrong with this information.
Lebanon operates under a strong semi-presidential system. This system is unique in that it grants the president wide unilateral discretion, does not make him accountable to Parliament, unless for treason, yet is elected by the Parliament. The president has the sole power to appoint the prime minister, and may dismiss them at any point, without input from the Chamber of Deputies, which can force the president to resign.[14]
The president has the sole authority to form a government (which must then receive a vote-of-confidence from Parliament) and dismiss it when they wish. This makes Lebanon a president-parliamentary system rather than a premier-presidential system, such as France, as the president does not have to cohabitate with a prime minister he dislikes. The historical reason for the broad powers of the president are that their powers were merged with those of the French high commissioner of Greater Lebanon, creating an exceptionally powerful presidency for semi-presidential systems.[15]
The first statement, that “Lebanon operates under a strong semi-presidential system” is definitionally incorrect. Although not all the political science work is consistent in how it defines a semi-presidential system, most of it requires that the president be elected popularly. As the next sentence of the first paragraph notes, correctly, the President of Lebanon is elected by parliament. The first work that I know of in English to use the term, semi-presidential, was an article by Duverger, published in 1980. He explicitly refers to an elected presidency with “considerable powers”; the latter condition may be vague, but the source of the office in popular election is clear. And it has been followed by most political scientists as best I can tell, even though there may be a few works here and there that (lamentably) use “semi-presidential” as if it meant no more than “hybrid” or “I do not know how to classify this one.”
As for the second paragraph, I suppose I should be happy that the author of this particular article is aware of the distinction between premier-presidential and president-parliamentary. This is a distinction first introduced to the literature by Shugart and Carey (1992). It has been adopted by several other authors, including the leading specialist on semi-presidentialism, the late Robert Elgie. The article gets the distinction between these (sub)types basically right, that president-parliamentary rules grant the president dismissal power over the government (premier and cabinet). However, there is just one problem: If the president is elected by parliament, the system can’t be premier-presidential or president-parliamentary, just as it is not semi-presidential.
Okay, so what is it? It is either a parliamentary system or perhaps some hybrid that is distinct from any form of semi-presidential system. The settling of this question might depend on whether the president really has the authorities this Wikipedia author ascribes to it. My reading of the Constitution of Lebanon (1926, with revisions through 2004) is that the president has no such powers. There is one potential ambiguity, but to me this is a fairly straightforward parliamentary structure. (At least if it were a democracy, which of course it is not.)
First of all, the president’s election is indeed in parliament. Article 49 states that the president is elected by the Chamber of Deputies by a two thirds vote “in the first round.” The goal of having the president be supported by super-majority is certainly more consistent with a presidency that is intended to be essentially ceremonial rather than an active political figure (as is likely to be the case with popular election). The constitution does not seem to stipulate what happens in subsequent rounds, if two thirds is not reached. Some other parliamentary constitutions have initial two-thirds requirements and then drop the requirement to a simple majority in later rounds because the goal of filling the post is more important than the goal of having a broadly supported head of state. Lebanon’s presidency has been vacant for about two years, so it may not be inadequate reading on my part; rather, there may actually be no procedure to fall back on when the first round does not produce a candidate who can win the votes of two thirds of the Deputies. (Reuters offered an explainer in 2022.)
As for powers, once in office, Art. 53.1 states that the president convenes the cabinet at any time “without participating in the vote.” Art. 53.3 says “he promulgates a decree appointing the prime minister separately.” In context, it is not clear what this process is “separate” from. In any case, a decree appointing the PM sounds like potentially real power, and it is a little ambiguous. However, from further context it seems this is likely a ministerial act, not discretion. (Lots of constitutions state that the president does such and such a thing “by decree” or similar act, but it is a formality.)
In Art. 54 we find that nearly all acts of the president must be counter-signed by the PM. So much for “wide unilateral discretion.”
Art. 18 states that the Chamber of Deputies and the Council of Ministers (cabinet) may introduce legislative bills; notably the president is not mentioned here. In Art. 56 is described a presidential veto that can be overridden by an absolute majority.
Art. 69.1.d states that the government must resign when a new president takes office. I think this is unusual in parliamentary systems, but does not take a case outside that category.
Finally, and very importantly, the procedures for forming a government (Art. 64.2) explicitly require that a new Council of Ministers submit its program of government to the Chamber of Deputies where it must secure a vote of confidence. Art. 69 lists the conditions under which a government is considered to have resigned. It describes typical parliamentary rules, such as a no confidence vote, but does not mention the president having discretion to dismiss. I see nothing that indicates that the president can dismiss a PM “at any point,” as the Wikipedia article states.
This reads as a quite straightforward parliamentary system to me. I do not doubt that there are informal understandings that provide the president (when there is one) with more influence than the constitution implies. I leave this to experts on the country’s politics. In fact, Lebanon famously parcels its top positions among representatives of its main sects. This process is not even mentioned in the constitution. But as to formal powers, this is a parliamentary system, and most certainly not a semi-presidential system or any subtype thereof.
Wikipedia has become a great source of factual information (like electoral results). But I caution readers not to rely on it for classification of regime types, electoral systems, and other political institutions. Come to this blog or other reliable (I hope!) sources for that.
The British Columbia provincial election was held 19 October. It was expected to be close and close indeed it is. Results show the New Democratic Party (NDP) narrowly ahead of the Conservatives, with votes 44.6% and 43.6% respectively. The provincial assembly has 93 seats, elected in single-seat districts by plurality (FPTP). At last check, the NDP is shown as having won 40 seats and to be leading in an additional 6. The Conservatives also have won 40, and are leading in another 5. The Greens will have 2 seats (2.15%) on 8.2% of the vote. Note that a majority is 47 seats, and thus it currently seems likely that there will be no majority party.
UPDATE: In the final count, one seat (Surrey-Guildford) shifted to the NDP with a margin of 27 votes. So, unless the recount changes this back again, the NDP will have won 47 seats and its leader has been formally asked to form government. The rest of this post is unmodified.
At least five districts (ridings) have margins between the NDP and Conservatives within a percentage point, and several more are within 1.5%. CBC says it could be a week or more before the result is declared, given those 11 seats where the lead is too close to call. B.C. election law requires an automatic recount if the margin of victory is 100 votes or less.1
The campaign was notable for the sudden withdrawal, on the 28th of August, of one of the three main parties. The party that left the contest was the B.C. United, successor to the B.C. Liberal Party. This party governed the province for several terms over the past quarter century and was the official opposition in the preceding term of parliament, but it had fallen to third place in this contest. It withdrew to avoid splitting the center-right vote, allowing that segment of the electorate to flow to the B.C. Conservative Party. The Conservatives had been a marginal force, getting only 1.9% in the previous (2020) election. Just like that, they may yet turn out to have won this election, or else they will be positioned as a strong opposition.
Below is the polling trend, lifted from Wikipedia. The top graph shows the campaign period, all of which was after the BCU withdrawal. The bottom panel has the graph since the previous election. Note how the BCU was clearly in second place till some time in summer 2023, when the Conservative vote started to surge, and eventually overtake BCU. By the time the BCU withdrew, it was already effectively a two-party race but with a high probability that vote-splitting on the right would have handed victory to an NDP winning less than 45% of the vote. The late polling trend shows some erosion of the Green vote after about the middle of 2023, as the race became more focused on NDP vs. Conservative.
So this is what those who favor a “Duverger’s law” explanation of plurality elections expect, right? It appears voters were already “coordinating” once it became clear that the Conservatives were a viable option on the right, and then elites (BCU leadership) followed suit by pulling out of the race. In the meantime, some BCU sitting legislators had defected to the Conservatives, starting with the member who became the leader of the latter party (John Rustad).2 Some BCU candidates, including five incumbent legislators, remained in the contest as independents.3
The effective number of vote-earning parties from this election is 2.52, and the smallest that it appears possible for the effective number of seat-winning parties to be is 2.05. That would be pretty Duvergerian! However, it is based on the surely unrealistic assumption that all 11 undeclared seats go to the same party (giving that party 51 seats–as each already won 40). If instead each party wins the seats in which it currently leads, we have a largest party with 46, and effective number of seat-winning parties of 2.09. More importantly, of course, that would be no majority party, as the largest would have 49.5%, one seat short.
By Taagepera’s Seat Product Model (SPM), an electoral system of FPTP in an assembly of 93 seats is expected to yield an effective number of seat-winning parties of 2.13, which would be slightly more fragmented than the actual result seems likely to be. However, the effective number will be so low in this election mainly due to the large size of the second party, which will have 48.4% if it wins the five where it is currently leading for a total of 45. The Seat Product Model expects the largest party in an assembly of 93 districts decided by plurality to have 56.7% of seats, which would be (rounding) 53 seats. I was curious what that would imply for three parties getting seats4–no other party that remained in the running reached even 1.5% although the various independents (many of them ex-BCU, as noted) combined for 2.3%.
To get an effective number near 2.13 with a largest party having 53 of the 93 seats would imply the third party–here the Greens–getting 5 seats (instead of 2) and the second party getting only 35. That works out to 2.1308. In other words, even a small assembly of just 93 seats should have room for a significant third-party presence, even if it also is expected to have a substantial majority for the largest party. The Greens have carved out a substantial niche for themselves despite the obstacle of FPTP. Moreover, contrary to the Duvergerian orthodoxy, having a third-party presence is quite the expected outcome! In fact, some third party “should” win a few more out of 93 districts. Perhaps a Green Party is especially poorly placed to be such a party, having a relatively narrow geographic reach in which it can appeal. The party did not come especially close to winning in any district other than its two success. In fact, maybe a “politics-blind” model like the SPM implies that the third party should be a regionally based right-wing party. In other words, what might have resulted had the BCU not collapsed and the Conservatives not been the only party on the entire right side of the spectrum.
This scenario of a regionally based party on the right winning some seats implies a system of four seat-winning parties, given that in actual BC politics a Green Party has carved out a foothold. Can we get to our expected effective number of seat-winning parties of 2.13 and largest party with 53 of 93 seats while leaving room for another party? Sure we can. For instance, NDP 53, BCU 35, Conservative 5, Green 2 gives us 2.137 (close enough). Something like that might have happened if the party system had stayed about where it was in spring or early summer of 2023. Or flip the two right-wing parties and it might describe a plausible result had the Conservative surge stalled where it was in early 2024, leading the BCU to tough it out rather than throw in the towel. Such an outcome would have been what Cox (1997) would refer to as a “non-Duvergerian” equilibrium, wherein it is not clear which of two contenders on one side of the ideological spectrum is better positioned to be the main alternative to the stronger party on the other side. However, as we can see from the graph above, the slippage of the BCU and the rise of the Conservatives occurred very rapidly. The period from October 2023 to March 2024 might have stabilized to where the BCU held steady in the 15–20% range and Conservatives under 30% (with NDP dipping to around 40%). Instead, the specific dynamics of this contest–a right motivated to defeat the incumbent NDP under condition of one of those parties experiencing a rapid surge–produced an outcome under which Duverger trumped Taagepera. It actually does happen. Sometimes.5 The bottom line is that elections are rarely this close6, and closeness matters!
It seems odd to me to specify the recount threshold by raw votes rather than a percentage. ↩︎
Rustad had been kicked out of the BCU caucus for questioning the science of climate change, or more to the point for urging people to “celebrate CO2.” ↩︎
As Ryan notes in a comment, former BCU candidates who remained in the contest as independents might have made a difference in Vernon-Lumby (Kevin Acton got 15.5% and the NDP incumbent is leading by 1.4) and Richmond-Steveston (Jackie Lee got 10.1% and the NDP incumbent is leading by 1.9). There was also the case of Adam Walker, an independent who received 16.4% of the vote in Ladysmith-Oceanside. As Ryan also notes in a comment, Walker was formerly a member of the NDP. This is a riding that nonetheless still went to the NDP with 41.2% with a Conservative on 35.8% (Green 6.6%). Walker would have needed considerably higher vote total to “spoil” this contest. ↩︎
We should pause and ask if three parties getting at least one seat is expected for a seat product of only 93, according to the SPM. Yes it is (the fourth root of 93 being 3.1). ↩︎
In case anyone is wondering whether the SPM is less successful at capturing FPTP systems than PR, no. Across all simple electoral systems, the mean ratio of the effective number of seat-winning parties to the number predicted from the product of mean district magnitude and assembly size is 1.07. For FPTP the mean is 1.02 whereas for PR it is 1.10 (a statistically insignificant difference). So the SPM, which for FPTP predicts solely from the number of districts, is even more accurate for FPTP than it is for PR. But it is pretty good for both! (Data referenced here are from national-level elections, of course. Total number of elections is 688, of which 273 are FPTP.) ↩︎
For national-level data, an election with a margin of one percentage point of the vote between the top two parties is at about the 94th percentile. For seat shares a margin of one percentage point is at the 98th percentile! ↩︎
There are various indications that a counter-strike of some sort is expected imminently from Israel against Iran in response to the latter government’s second major barrage of missiles and other projectiles on Israel. For instance, Iran has closed its airports. I refer you to the post I wrote way back in early August. A lot of very significant events have happened since then, but nothing in the thesis I articulated then has changed.
And, yes, the programming note remains in effect, one year later, and likely will so remain for some time.
(An hour or so after I posted this, I read that Tehran’s main airport reopened. So who knows…)
The German state of Brandenburg held its general election on 22 September. Like most other German states–including the two that had voted three weeks prior–and the elected chamber (Bundestag) of the federal parliament, Brandenburg uses a mixed-member proportional (MMP) system.1 Brandenburg’s MMP system consists of 44 seats elected in single-seat districts and 44 from a compensatory party-list tier. The far-right Alternative for Germany (AfD) did well, almost 30% of the vote, although it came in second place. The plurality of votes and seats was won by the Social Democratic Party (SPD), the party leading the incumbent federal coalition (formed following the 2021 election).
The election result is particularly notable for the high share of votes that fell below the 5% party-list threshold. In fact, only four parties cleared the threshold and won any seats, while a total of five parties had over 4% of the list votes and three more had at least 2%. The total vote for parties below the threshold amounted to 14.3%. That is a lot of wasted votes, and this included one of the coalition partners in the federal government, the Free Democrats (FDP) who had an embarrassing 0.83% (down from just over 4% in the preceding election; in other words, they did not win seats then, either, but were at least competitive). Another national party that used to be strong in the eastern part of the country where Brandenburg is located, The Left, fell to just under 3% (from 10.7% in the previous election).
The only parties that will be represented in the 88-seat state assembly are: SPD, AfD, the Sahra Wagenknecht Alliance,2 and the Christian Democratic Union. These parties’ votes (and seats) percentages are as follows: 30.9 (36.4), 29.2 (34.1), 13.5 (15.9), 12.1 (13.6). Note that each party is significantly over-represented, which is of course the direct product of having over 14% of votes not count towards any seats. The effective number of seat-winning parties is 3.42; the effective number on votes is 4.60 (using list votes).
Only two parties won any constituency seats, and in this election the nominal tier of single-seat plurality contests produced a plurality reversal: SPD won 19 seats on 33.6% of the nominal votes, but the AfD won 25 on 31.5%. Given the fully compensatory nature of the allocation of party-list seats under MMP, the SPD was allotted 13 list seats to bring it up to its proportional entitlement (proportional to each party’s share of above-threshold list votes, that is), while the AfD won only 5 such seats. The other two parties that won representation won only list seats–perfectly demonstrating how compensation from the list works under MMP.
The most likely governing coalition will be SPD + Sahra Wagenknecht, which combines for 46 seats (52.3%). This is the only combination of two parties not including the AfD that surpasses 50% of seats. Note that it would be a manufactured majority, as the two parties combined for not quite 44.4% of the votes. The more “conventional” combination of SPD and CDU is only 44 seats, which happens to be precisely half the seats (on 43% of the votes).3
Outcomes like this manufactured majority and challenging coalition formation combinations are going to become more and more common in Germany as long as vote fragmentation remains high and as long as the AfD continues to advance into the position of one of the leading parties (and as long as other parties refuse to work with it).
Arguably, the 5% threshold is no longer serving Germany, including states like Brandenburg, well. The high threshold, in the context of such vote fragmentation, is producing a very high wasted vote, as this election shows. It is leaving otherwise viable and more ideologically moderate coalition partners out of seats entirely. For instance, if the threshold were 3%, the Greens would have won seats even without assuming any changes in voting behavior from a lower threshold. In such a case, probably SPD+CDU+Greens would have had a majority of seats (albeit still short of a majority of votes, at about 47%). A threshold of 2% would have included The Left among the seat winners as well. I am not normally a fan of changing rules based on very specific effects they would have on certain parties. However, it is noteworthy that the threshold was originally justified as a way to keep extremists out, yet it fails in that objective when extreme parties become among the biggest parties in the system whereas some relatively more moderate parties are excluded. Perhaps it is time for a serious debate in Germany over whether 5% is too high a threshold.
See the post on the earlier state elections for some links to information on this new party, which clearly has attracted voters from The Left, among others. ↩︎
If the assembly had 87 or 89 seats, these two might have had a majority, although to be sure one would need to run the electoral formula and determine which parties had the last quotients. ↩︎
Fruits and Votes 