South Korea 2024

South Korea held its National Assembly election on 10 April. The result was a strong opposition majority. Given that this was approximately a midterm election, held at about 42% of the president’s term, a strong result for the opposition should be expected from standard electoral cycle effects in presidential systems.

The president is elected for five years, and the last election was in March, 2022. The assembly is elected for four years. Thus the last presidential election was held about two years into the assembly term; it, too, produced an opposition win, with the minority in the assembly seeing its candidate win the presidency, Yoon Suk Yeol of what is now called the People Power Party. (Its main component was known as United Future in 2020.) The 2022 election was close, with Yoon winning 48.6% and the Democrats’ candidate, Lee Jae-myung, winning 47.8%. The Korean presidency is elected in one round, by plurality. Plurality election of the presidency combined with nonconurrent 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. Since 2020, that system has been designed to be more proportional than had been the case for previous elections. But as we shall see, it has not worked that way in practice.

The electoral system, as best I understand it, is a combination of MMP and MMM.¹ In addition to the single-seat districts, decided by plurality, there are two sets of list seats, totaling 47 seats. Of these, thirty are elected as compensatory seats (allocated from list votes after taking into account districts won by each party, as under MMP) and 17 are parallel (added on based on proportions of list votes without regard for district results). This was already the system adopted for the 2020 election. If this helpful Twitter thread by Raphael Rashid got it right, then the system seems indeed to be the same as in 2020. In prior elections, the system had a purely MMM design.

Where I could use some help is figuring out the detailed outcome by party and bloc. The table at the Wikipedia page for the election is less helpful than usual. Below is the main part of it as a screenshot. The part I am reproducing should include all the parties that won seats.² There are many more parties in the original that won no representation.

Rashid’s thread indicates that the two main parties each formed a “satellite” party to exploit the compensatory provision. This sort of behavior by parties has occurred before under MMP³ and is a well known pitfall of the system, if the electoral law or election administrators do not take steps to counter it. A party can win its nominal-tier (single-seat district) seats as usual but not present a party list, while setting up a formally separate party that competes only for list seats. In doing so, its list wing (the satellite) would get its full proportional share of the compensatory list seats without having to deduct the nominal seats won by the main party, given that they are formally separate parties. This was done in 2020 in Korea as well, although as best I could tell it did not affect the majority won by the Democratic Party in that election.

Rashid indicates that for the 2024 election the Democratic Party created a Democratic Union as its satellite while the People Power Party (President Yoon’s party) created a People’s Future as its satellite.

Unfortunately, the Wiki table does not break these out. In fact, it does not even indicate anything called Democratic Union. We see that various partners within “Democratic Alliance”⁴ won list (“proportional”) and nominal (“constituency”) seats, or both, and that their total is 176 seats, of which 14 were list seats. The combined list votes of these parties is only 26.7% of the total, whereas 176 seats is 58.7% of the total 300. So this certainly has an MMM (parallel) feel to it. In fact, winning a large majority on barely a quarter of the votes is a rather extreme manifestation of disproportionality! (Assuming these numbers are correct; I note the table shows no nominal votes, even though voters get both a nominal and a list vote.)

Those 14 list seats for the various parties of the Democratic Alliance would be about 29.8% of the total 47 list seats. This is again quite consistent with how an allocation would work under a fully MMM system, in that the list seats–as distinct from the seats overall–are themselves roughly proportional to the list-vote share. And that is exactly what the satellite-party strategy does: it effectively converts MMP to MMM for a group of “parties” pursuing this strategy. Note that if all 47 list seats were allocated in a compensatory matter, and the Democratic Alliance were treated as a single party for purposes of calculating seats, it would have won zero list seats, given that its number of nominal seats (districts) won was already (far) above its proportional entitlement, at 162 (54%).

The Wiki table combines People Power and People Future. It shows 36.7% of list votes for this combine, with it winning 18 list seats. That is 38.3%, again almost proportional to the list votes, and these seats are added on to the 90 nominal seats won. Again, just like MMM.

So the satellite-party behavior clearly achieved its intended result, by circumventing the compensatory feature of the 283 seats supposedly allocated via MMP (253 constituencies plus 30 list seats). I wish the breakdown included how many of the two types of list seats each party won (i.e., separating the 17 parallel seats from the 30 supposedly compensatory), as well as breaking down the “real” and “satellite” parties within the blocs. Maybe someone can point me towards an analysis that did so. I tried to locate something of this sort and did not turn it up.

South Korean democracy would probably be better served by a more proportional system. Or by proper enforcement somehow of the intended compensatory nature of the current system (although just ten percent of seats supposedly allocated in that manner is pretty tepid proportionality). It would also perhaps be served better by concurrent elections and a presidential election method other than plurality. Various sources report some degree of backsliding in Korean democracy in recent years, and while possible institutional reforms I mention here would surely not be sufficient to halt such trends, they just might help.

1. Mixed-member proportional and mixed-member majoritarian, respectively. ↩︎
2. For some reason the table shows 46 list seats and 254 constituency seats, but this seems to be incorrect. In any case, this small discrepancy barely changes the analysis and puzzles I am writing about. I also wonder why the table indicates two “independents” having won via party lists. ↩︎
3. See past plantings on Lesotho, and in particular the interesting discussion in the comment thread on MMP manipulation in that country in 2007. It also happened in Albania, and is the main reason why that country became an ex-mixed member case. ↩︎
4. Likely just a translation variant, but even so, the grouping in the table does not clearly indicate the satellite phenomenon. ↩︎

Vanuatu referendum on party-switching ban

The following entry is contributed by Henry Schlechta. 

Vanuatu will conduct its first-ever referendum on 29 May, on two proposed constitutional amendments regarding party-switching.

The first proposed amendment (Article 17A) provides that if a Member of Parliament resigns from a party, or is removed from that party under that party’s constitution, their seat will become vacant. The second proposed amendment (Article 17B) provides that an independent Member or a Member elected as the only member of a party must join a party with more than one member within three months of being elected: they would then become subject to the provisions of 17A, if that amendment is approved.

The full text of the amendment is available here. As passed by Parliament, the amendment also included a prohibition on no-confidence votes twelve months after a Prime Minister is elected. This provision is not required to be subjected to a referendum, and is presumably already a part of Vanuatu’s Constitution.

As I’ve posted about before, Vanuatu’s single non-transferable vote (SNTV) electoral system has tended to produce highly fragmented parliaments, with the 2022 snap election proving no exception; seventeen separate parties won at least one seat, with the largest two parties winning only seven seats each, giving an effective number of political parties (in terms of seats) of 12.2. Since that election, Vanuatu has had three Prime Ministers and a brief constitutional crisis surrounding a no-confidence vote.

Against this backdrop, it’s unsurprising that constitutional amendments to reduce party-hopping and no confidence votes would be favoured. These amendments, which came out of a constitutional review process from 2016 (I have not found a copy of the report from this process, if anyone has seen it let me know in the comments), seem to have passed Parliament unanimously, and based on a cursory look at social media the Yes campaign seems to be well organised.

However, at the same time, it seems possible that the amendments will fall short of their promise. Neither the party-hopping changes nor the ban on no-confidence votes strikes at the root cause of instability – the highly fragmented legislatures. Governments will still need the support of a wide constellation of parties, even if those parties will hold greater control over their legislators. The twelve-month period where no-confidence votes are suspended will likely have more effect, but even in countries like Papua New Guinea where such provisions are in effect the underlying fragmentation still creates political instability.

Obligatory threshold provisions for European Parliament, 2024

Elections for the European Parliament will be held in early June. The provisions for the elections now require thresholds in high-magnitude electoral districts.

All Member States must use a system based on proportional representation. In addition to the voluntary threshold for the allocation of seats of up to 5% at national level, Council Decision (EU, Euratom) 2018/994 established an obligatory minimum threshold of between 2% and 5% for constituencies (including single-constituency Member States) with more than 35 seats. This requirement must be met in time for the 2024 European elections at the latest.

Currently, the following Member States apply thresholds: France, Belgium, Lithuania, Poland, Slovakia, Czechia, Romania, Croatia, Latvia and Hungary (5%); Austria, Italy and Sweden (4%); Greece (3%); and Cyprus (1.8%). The other Member States apply no threshold, although Germany tried to do so, but in two decisions of 2011 and 2014, the German Constitutional Court declared the country’s existing thresholds for EU elections (which were first 5%, then 3%) to be unconstitutional.

Most European member states use a single statewide district. These range in magnitude from 6 (Cyprus, Luxembourg, and Malta) to 96 (Germany¹). The others that have district magnitude of 35 or greater are Spain (61) and France (81). Aside from the case of Germany mentioned in the quote above, Spain is the only remaining state that did not already have a legal threshold, or one in the stipulated range.²

The mean magnitude of the single-constituency cases is 24. Four states divide their territory into multiple constituencies: Belgium, Ireland, Italy and Poland. The mean district magnitude for this latter group of countries is just under 8. Overall, there will be 47 districts for 720 seats, making a mean district magnitude of 15.3 and thus a seat product of 11,016.³

(Please note various clarifications and corrections offered by readers in the comments section.)

1. There is an interesting twist on the single German district: “Seats are apportioned [sic–“allocated” would be better] to parties nationally. A party can choose to only stand in some of the 16 states and have its national seat count be subapportioned to those states. Only the CDU and the CSU have done this in previous elections.” (Quoted from a footnote at Wikipedia.) Also, if I understand other footnotes correctly, the seats parties win in Italy and Poland are determined nationally, despite their being “districts”; apparently a two-tier compensatory system. This is a level of complexity that calls into question the mean magnitudes I report below, but I am not going to attempt to adjust for this factor. ↩︎
2. It is interesting that Spain had no threshold despite the high magnitude. After all, Spain applies a district-level legal threshold (3%) in elections to its national parliament, which has real impact only in the two districts that have 30 or more seats. ↩︎
3. Whether the seat product model “works” or should even be expected to work for a supra-national parliament is not something I have ever entertained. ↩︎

RCV in a Republican presidential primary? Meh

I subscribe to email alerts from various organizations within the US electoral reform movement. One of these organizations sent me an email last week that said, in part:

Yesterday, the Republican Party in the Virgin Islands became the first Republican state or territorial party to use ranked choice voting (RCV) in its presidential primary!

Please pardon me if I am not quite so excited. Not because it is the Virgin Islands and only four delegates were at stake. But because the way RCV was used is hardly a significant advance, and was quite irrelevant given the nature of the contest in the territory. (NB: It was a caucus, not a primary, but that does not change any of the argument I am making here, and probably would not temper the organization’s excitement.)

The ranked choices are used to determine a single winner. Thus the Virgin Islands GOP is using a “winner take all” system despite there being multiple delegates. This is hardly the sort of advance I would expect a pro-preform organization to be trumpeting. 

The email further told me that this procedure ensured that, unlike in the New Hampshire primary earlier in the month, this RCV primary meant votes were not wasted. It allowed ‘voters to decide among the full field, instead of party elites pushing candidates out lest they “play spoiler” or “split the vote.”’

With multiple delegates at stake and more than one candidate, arguably votes not cast for the front-runner were indeed wasted, as that candidate got 100% of the delegates despite being well below 100% of the votes. Further, if the objective is to keep competitors in the contest longer,¹ there is an obviously superior solution: allocate the delegates proportionally. This is the procedure used, with large reservations, in all Democratic Party presidential primaries and in some states by the Republican Party. If the GOP contest this year were in a setting in which every delegate counted (like the Democrats in 2008 or at the earlier phases in 2020), it might be good to let someone other than the front-runner come out of the Virgin Islands, or any other contest, with a delegate or two. Instead, the front-runner got all four. 

Moreover, that front runner won two thirds of the (first-preference) votes. In other words, the transferring was quite irrelevant in this specific contest, given that it is a winner-take-all rule in force. By having over half the votes, he was going to get all four delegates regardless of anyone’s second choices.

Had proportional representation (PR) been in place, the vote distribution likely would have meant the front-runner getting three of the four, as–depending on the specific rule–Nikki Haley’s 20.6% should have been sufficient for one of the four.

In the actual transferring from distantly placed candidates, Haley gained 12 votes and the front-runner gained 8. (This being a caucus with evidently rather less excitement on the ground than in certain reform circles, there were only 253 total votes cast, 246 of which remained with choices marked to the end.²)

Another way RCV could be used–which would be similar to how a few Democratic Party contests now work–would be to use eliminations and transfers only for (first-preference) votes cast below the share needed³ to win a delegate. In the initial count, for example, Ron DeSantis started with 5.9% (15 votes), even though he had already suspended his campaign. His voters could not have allowed Haley to reach two delegates even if the four available were allocated proportionally after transferring non-viable candidates’ votes. However, because she ended just over a quarter of the votes after full transfers, the final “two-candidate preferred” surely vote would have guaranteed a delegate for Haley even if the specific rule in place had not been favorable to her winning one delegate on her first preference votes.⁴

Such use of RCV in a presidential nominating primary or caucus would be far superior to simply using it to determine who gets all the delegates. I will concede that if the front-runner had well under 50%, RCV might hand the full slate of delegates to someone other than the plurality candidate, demonstrating how “voters can identify a consensus, majority nominee in a multi-candidate field,” in the phrasing of the organization’s email. But in such a case, given multiple delegates at stake, would it not be that proportional allocation (whether or not combined with RCV) is obviously even more superior? One would think so. But people will get excited over whatever floats their boat, and some pro-reform organizations have developed something of a fetish over RCV.

Finally, it is kind of ironic, but the Republican Party in the Virgin Islands is actually violating party rules by holding a winner-take-all event prior to March 15!⁵ But, hey, they used RCV!!!

1. Never mind that by the time this contest happened, it was well publicized that only two candidates were still active. ↩︎
2. As of December, there were 2,107 registered Republicans in the territory. That is a turnout of around 12%. ↩︎
3. Whether due to being too low to qualify based on the proportional formula in use, or due to a legal threshold (e.g., 15% in Democratic Party contests). ↩︎
4. The delegate primaries that use “proportional” allocation in US currently often use non-standard rules. What if, for example, the requirement were a full Hare quota (1/4 of votes)? Then Haley would have won a delegate only if preferences for other candidates aside from the top two were transferred before making the delegate allocation. ↩︎
5. The penalty for this is that they got just those four delegates to the convention whereas it otherwise would have been nine. ↩︎

Electoral systems and clientelism

That title could be the title of a book, or a multi-volume collection. This is a big topic. The purpose here will be more modest than the title implies. I want to sketch the basic logic of how electoral systems can generate incentives for “clientelism.” I responded already to an excellent question in the thread on proposed electoral reform in Malta. In the editorial that I quoted in part, it is claimed that the use of the single transferable vote (STV) in Malta has led to clientelistic behavior by which politicians “rely on dishing out favours to secure their seats.” The money quote on why the electoral system promotes this behavior would be the following:

The STV system, while aiming to provide representation to diverse political views, has led to fierce competition among candidates from the same party within a single district. This unhealthy competition often results in politicians prioritising convincing voters not to vote for their internal rivals rather than engaging with people from opposing parties.

Christopher asked, “Is there a particular tendency towards clientelism in STV and SNTV¹ that is not present in open list PR?” This is a great question. I answered it in that comment thread, but I felt it deserved its own dedicated planting, as well as further elaboration on the broader issue.

First, let’s consider a working definition of clientelism. It is, very broadly, the exchange of favors for political support. Not mere campaign promises, and not routine casework (like a legislator helping a constituent navigate the bureaucracy). But actual delivery of favors as a quid-pro-quo for the vote. It is most prevalent when vote choices can actually be observed, but such monitoring is not required. Parties or candidates can have other means of knowing which voters are likely to be loyal, and direct favors to them without actually validating how they voted. (They typically can validate that they voted, and have other good information on their political allegiances.)

Clientelism is also not “pork barrel” politics. “Pork” is local public goods, such as a road or train station or a military base (etc.) being located in a constituency as a reward for its voters being politically loyal to the party or legislator. Pork is not mainly directed at specific voters or narrow groups of voters. These two concepts can shade into one another, however, without a sharp dividing line between them. Nonetheless, they are conceptually distinct.

The most pernicious forms of clientelism are those that involve the subversion of public services. Suppose there is some entitlement program such that anyone who meets program criteria should get it (a health or childcare benefit or an electrical grid connection to your house, for example). Yet in practice, only those who prove their political loyalty actually get the benefit. If the local clientelistic “boss” does not vouch for you, too bad. No service for you. The “boss” could be a local party worker, a government official who is part of the clientelistic network (and a patronage appointee rather than one hired principally on merit), or even our legislator.

There are other, more “petty,” forms of clientelism, such as legislators showing up to weddings and funerals and bringing gifts as a means to cement political loyalties. These practices cost money, which may be raised through unregulated donations from business or via outright corruption,² although they may also be financed through more legitimate campaign contributions. But they do not undermine public services like the first kind I mentioned. Basically, clientelism takes many forms, and some may be very harmful to effective governance and some not so much.

There are also practices to cultivate a “personal vote”³ that are not clientelism at all. Candidates emphasizing how their own attributes–like being locally based, being of a certain occupation or other experience, etc.–will make them better representatives than their competitors are not clientelism because nothing of intrinsic value is being exchanged.⁴ 

Having laid out these broad definitions and boundaries of the concept, let’s turn to what the electoral system may have to do with it. Back to Christopher’s question–is there a particular tendency towards clientelism in STV and SNTV that is not present in open list PR? On theoretical grounds, the answer to the question is yes.

An inability, given electoral rules, of parties to maximize their seats without engaging in vote management is what makes it critical to give voters a reason to favor one candidate of the party over another. Clientelism is one–but only one–means of managing the vote, by which here I mainly mean attempting to ensure each candidate the party seeks to elect has the personal vote totals needed to earn a seat. Any number of personal vote-enhancing tactics or attributes might work, but the advantage (to the party) of using clientelistic practices is that it gives voters a clearer incentive to vote for the “correct” candidate because that is the individual who got you, the voter, the favors.⁵ (Vote management also may take place at the candidate selection stage,⁶ but I will mostly leave that aside here.)

The imperative to engage in vote management is most acute under SNTV, somewhat less under STV, and theoretically a non-factor under OLPR. With SNTV, by definition the winners are the candidates with the M highest individual vote totals in the district, where M is the magnitude of the district (the number elected from it). Because there is only a single vote and it does not transfer (no ranked preferences) or pool (no party-level aggregation before seats are assigned), the party can win its target number of seats only by having that number of candidates within the top M. If it has too many candidates, or too many votes concentrated on its most popular candidate, it may fail to elect the number it is seeking to elect, with other candidates falling out of the top M. With STV, the vote transferring should limit this problem, but it does not necessarily eliminate it. If a party has some of the candidates it is hoping to elect get eliminated early in the count, the party will fail to elect all that they could have if only those candidates had stayed in the count long enough to pick up transferred preferences from more popular co-partisans or from eliminated candidates of their own or other parties. This is especially a problem if voters load a lot of their first preferences on one candidate (similar to one of the key problems in SNTV), and if they do not rank all available candidates of the same party before ranking candidates of other parties. (Failure to keep preferences within the party is known as “vote leakage.”)

Why, then, is the vote-management imperative that may drive clientelism a non-factor under open lists, at least theoretically? Because vote pooling at party-list level means the party won’t displace seats to another party by failing to have an optimal distribution of votes among its candidates. A list wins some number of seats, which we will denote as s, based on its collective vote total. It does not matter if some of the party’s s candidates with the top individual vote totals do not place in the district’s top M overall. It will always elect s, the number its pooled vote total entitled it to, as long as it has at least that many candidates on the list.

There can be other reasons why the party might care which candidates win, however, which means there can still be party-driven clientelistic strategies, or other vote-management practices, under OLPR. That is, while parties have incentive to nominate a full list of candidates to collect votes that pool for seat-maximization, maybe they really want to ensure the election of some specific candidates.⁷ So the clientelistic way to do this is give them access to resources to ensure their own election. (Again, there may be ways to promote a division of the vote that is optimal from the party’s perspective without using clientelism.)

That is the theory.⁸ When we turn to empirics, it gets murkier. 

An important point here is that clientelism usually is not caused by the electoral system. Clientelism can occur under any electoral system, and in authoritarian as well as democratic systems. It has many roots aside from institutions.⁹ It just takes different forms under different institutions. In theory, closed-list PR should be the electoral system least conducive to clientelism. CLPR should encourage campaigns centered around policy and programmatic offerings to voters, because voters select the party as a whole and thus candidates are not pitted against one another in their quest for a seat. Nonetheless, there is no doubt that parties as a whole can engage in clientelism. There are closed-list systems that are famous for it, such as Argentina¹⁰ and Venezuela (prior to the adoption of MMP in 1993, but clientelism certainly continued afterwards). This is a different form of clientelism, in that it is obviously not being used as a strategy to divide the vote within a party, as it would be under SNTV, STV, or potentially OLPR. But if you are a citizen who is unable to get a public service without having proven you have the “correct” political allegiance, you do not really care how the electoral system might be, or might not be, responsible for your predicament.

Returning to the prevalence of clientelism under systems with intra-party competition, we might note that Brazilian politics is clientelistic under OLPR, but Finnish politics is individualistic without being clientelistic despite also using OLPR. Maltese politics is clientelistic under STV, Irish much less so. (Ireland is famous for the localism of its politics and for “friends and family” voting, but I don’t think of that system as clientelistic–but see exchange with Ken in the comments.) Taiwan probably had less clientelism than Japan⁹ under SNTV and both surely had far less than Colombia, which used a system that was basically SNTV prior to 2003. 

Malta, as we were discussing at the other planting, has clientelism with STV despite a very party-oriented electorate. That suggests it is probably not vote-management, as I defined it above, that is driving the clientelism. So maybe STV is not driving it, either, but I do not know enough about Maltese politics to make such a judgment. It would be important to nail down the logic of clientelism in Malta before recommending an electoral reform. If it is more party-level clientelism, then closed-list PR will not help. The quote from the editorial (see above or the earlier post) certainly implies it is candidate-based, however.¹¹ [This footnote is actually no. 12 in list below.]

In sum, clientelism is the exchange of favors to individual voters or small groups of people in return for political allegiance, including votes. Such practices should be kept distinct from other means of campaigning on benefits other than those derived from programmatic policies, such as “pork” and are distinct from the personal vote. However, in systems that promote a personal vote via intra-party competition, parties may engage in various tactics of vote-management in order to maximize seats. Vote management may involve clientelism, although it does not inevitably do so. Clientelism can occur under any electoral system, but systems that promote vote management can be expected to encourage clientelism to a greater degree than other systems, all else equal. Open-list PR, because it does not usually tie party seat maximization to vote management, should not promote clientelism to the same substantial degree as SNTV does. STV should fall closer to SNTV than to OLPR in its clientelistic tendencies. Other factors matter; it would be a big mistake to think that the electoral system determines whether or not there is clientelism, but it may encourage it to take certain forms or be worse, in a given context, than would be the case with a different electoral system.

Footnotes. [Sometimes the numbering of these by Word Press goes awry.]

1. Single non-transferable vote. ↩︎
2. Perhaps because I mentioned it, I also have to define corruption. I hesitate to do so, but it is important to note that something is not corrupt just because it looks bad, and pork barrel is not per se corruption. A practice is corrupt if it involves illegal use of public resources, such as providing services only with direct payment (bribes), awarding public contracts to political supporters outside of regular bidding processes, or raising campaign funds illegally. ↩︎
3. A personal vote is that portion of a candidate’s vote that derives from his or her attributes or experience, as distinct from votes earned via the candidate’s party label. The classic definition is from Cain, Ferejohn, and Fiorina (1987). ↩︎
4. Engaging in casework, providing clientelism, and having a reputation for delivering pork all may be ways to build a personal vote. But a personal vote need not be built on any of these strategies. It could be as simple as voters liking the candidate who is of their ethnic group, resides in their area, is a celebrity (i.e., famous before entering politics), or whose personal integrity they admire. ↩︎
5. Incentives can be aligned in other ways, such as through party personnel practices. Examples include assigning legislators who will serve as conduits of regulatory and other favors on behalf of particular business interests to specific legislative (or internal party) committees where they can credibly claim to have lobbied their party or the government on behalf of the interest. This is not necessarily clientelism (or pork), but it may be, and certainly was in Japan under SNTV. ↩︎
6. For instance, by nominating/endorsing only as many candidates as the party perceives as optimal in a system without party-level vote pooling (like SNTV), or by ensuring a set of candidates with an appropriate mix of personal vote-earning attributes that will appeal to sub-constituencies within the district. ↩︎
7. This imperative becomes even greater if the list is comprised of an alliance of two or more parties. In that case, the component parties obviously are not indifferent as to which specific candidates win the seats the list obtains. This provides a strong SNTV-style dynamic to the competition. Alliance lists are common in the open-list systems of Brazil, Chile, and Finland. Chapter 14 of Votes from Seats deals extensively with this matter. Even a single party may have a collective interest in ensuring some specific candidates on its list get elected, perhaps for the “party personnel” reasons mentioned in an earlier footnote (i.e., wanting personnel with certain experience to serve as policy specialists on behalf of key constituencies). ↩︎
8. These theoretical points are advanced in more detail in chapters 13 and 14 of Votes from Seats, and also in my article with Bergman and Watt. Both also are empirical works, but the outcome of interest is the number of party-endorsed candidates and their vote shares, not clientelism or personal-vote earning strategies. On the theory, see also Carey and Shugart (1995). ↩︎
9. The most obvious roots would be economic, particularly uneven development. Richer and less unequal countries undoubtedly have less clientelism–at least of the “pernicious” forms. State capacity is also a big one, but the arrows can point both ways. Low state capacity creates more demand for clientelism to fill in gaps in public services, but clientelism creates vested political interests against building state agencies empowered to deliver services to the broader public without the quid-pro-quo. Ethnically divided societies also tend to be especially conductive to clientelistic strategies, mostly because ethnic affinity provides a “shortcut” to determining which parties or candidates are credible as service-providers to your group (see Chandra 2004). There are vast literatures on these points, which I will not attempt to review or synthesize here. ↩︎
10. On Argentina, and for an excellent overview of how clientelism works generally, see Auyero (2000). ↩︎
11. [Footnote marker no. 9, re Japan.] Most of the clientelism in Japan was of the more “petty” kind I referred to above. Campaigns were famously expensive and this fueled corruption. Personal vote and pork barrel were prevalent, as well, but not (as far as I know) the kind of subversion of public services that I think of as the most pernicious manifestations of clientelism. ↩︎
12. I struggle to some degree with understanding the full logic of this claim regarding Malta, however. With transferable votes, the reason for vote-management should be mainly driven by concerns of “vote leakage” whereby transfers do not stay within the party of the first-preference candidate. However, in Malta there is hardly any leakage; few votes transfer from one party to another (see the source cited in the post on Malta). In this context, personal-vote practices, whether clientelistic or otherwise, are not needed to maximize seats. It should be, in practice, more like what I said about open lists. I am certainly in no position to resolve this puzzle. ↩︎

Editorial in Times of Malta: Shift to nationwide list PR

The Times of Malta has run an editorial under the headline, “Scrap electoral districts.” It speaks favorably of a proposal to shift elections in the country from (modified) single transferable vote (STV) in thirteen districts, each electing 5 members, to a nationwide party-list PR system. (I will explain the modifications to STV briefly below.) 

Reasons offered in the editorial include both concerns over biased drawing of electoral boundaries and the clientelism of candidates seeking preference votes and competing with co-partisans:

First, it removes any incentive for the government of the day to meddle with district boundaries to suit itself.  The primary benefit of chucking electoral districts in the political dustbin, however, is that of making it much harder for candidates to rely on dishing out favours to secure their seats.

The editorial alleges that the clientelism has become worse.

Malta is increasingly turning nepotism, clientelism and tribalism into an art form. In a 2019 paper, political scientist Wouter Veenendaal used Malta as a case study of how smallness fosters clientelism. His analysis reveals not only that patron-client linkages are a ubiquitous feature of political life in Malta, but also that the smallness of Malta strongly affects the functioning of clientelism by eliminating the need for brokers and enhancing the power of clients versus patrons. Clientelism is found to be related to several other characteristics of Maltese politics, among which the sharp polarisation between parties, profound executive dominance and the incidence of corruption.

The STV system, while aiming to provide representation to diverse political views, has led to fierce competition among candidates from the same party within a single district. This unhealthy competition often results in politicians prioritising convincing voters not to vote for their internal rivals rather than engaging with people from opposing parties. [Emphasis added]

It is an interesting claim that clientelism is made worse by party polarization, and it may be so. If party lines are rigid and there are few voters willing to cross party lines in their ranking of preferences, then the candidates’ incentives are entirely to emphasize what they can do for their voters that their co-partisan candidates in the district can’t do. The arguments about size and clientelism are well understood in the literature.

With a country as small as Malta, and if the problems occasioned by intra-party competition are as bad as claimed, then both a single nationwide district and a party-list system (presumably closed lists¹) potentially make sense. However, by going to the large seat product² implied by a national district, they may get more party fragmentation than they want. Perhaps they should consider keeping districts, but fewer of them and thus higher magnitude, with or without a compensation tier. (Some examples of models are discussed in a footnote below and in the comments.)

Some more conventional PR system (whether districted or nationwide) certainly would be better than some other proposed reforms. Former PM Robert Abela suggested reducing the number of districts, but it is not clear if his ideas include a change from STV. If there were fewer districts–and assuming no significant reduction in assembly size³–then district magnitude would have to be larger. While that would help other parties aside from the main two (also being discussed as a goal of reform⁴), it also increases the intra-party competition and the information burden on voters to rank candidates.

The second-linked article mentions other proposals, including this odd one:

Minister Michael Falzon and Nationalist MP Herman Schiavone in January [2021] said it was time to discuss the possibility of introducing a national threshold to enable a third party to be represented in parliament, rather than having its votes wasted if none of its candidates were elected from a district.

A national threshold would allocate seats in parliament to a party that achieves a certain number of votes, say 5 per cent nationally, even if none of its candidates were directly elected from a district. 

Well, that is not what a “national threshold” means. One might presume that a potential nationwide PR system would indeed have a threshold. But if you are creating a mechanism to represent parties that earn some significant voter support do not win a district seat, you are talking about a compensation (or “top-up”) system, which indeed could also have a threshold.

I mentioned above that the Maltese system was “modified” STV. From the voters’ perspective, it is ordinary STV–they rank candidates and are not bound to stay within party lines in their rankings (thought most voters in Malta do⁵). And the process of determining district winners follows standard STV allocation practice. However, in a major departure from STV, there has been since 1987 a national determination of which party is entitled to a parliamentary majority. Adjustment seats would be added to parliament if the party with the majority of first preferences does not win a majority of seats across all districts, and this provision has affected election outcomes on multiple occasions. Then in 1996 this provision was extended to the party winning a plurality of first-preference votes, not only a majority. Thus a first-preference vote is not simply an indicator of which candidate a voter prefers–as by definition under STV–but is also taken directly as a vote for which party should have a seat majority (and hence govern). In 2007 this provision was further modified to ensure that the margin between the top two parties is reflected in the aggregate result.⁶ In other words, in effect, Malta already has a nationwide calculation of its seat balance in a sort of two-tier system, and has not been purely STV for a long time, but especially not since 2007. It therefore seems Malta has abandoned the STV principle already, and it may be time to shift to some sort of system that is actually designed to reflect nationwide party support and not merely “adjust” the district-level result.

(H/t to Jack Santucci for the Times of Malta editorial.)

[Note: I am aware there are some formatting issues. The Word Press editor has its issues, especially lately. It is hard to fix because on my screen it all looks okay. But on some devices, apparently not. Bear with me. Sometimes plantings in the orchard do not grow the way the orchardist intends. But we hope the fruit remains good.]

1. Or some sort of flexible list. A purely open list with a magnitude of 65 (or more) is not such a fine idea. ↩︎
2. The seat product for a single-tier PR (or FPTP) system is the mean district magnitude, times the assembly size. The Seat Product Model derives predictions from the seat product for expected party-system outcomes. See Votes from Seats. ↩︎
3. Current assembly size is 65, but can go higher due to provisions I discuss below. By cube root law, based on a population now around 537,000, we would expect it to be more like 81. ↩︎
4. Malta has a very pure two-party system, with third parties or independents winning any seats only very sporadically. With a seat product (not counting adjustment seats discussed below) of 325, it should be expected to have an effective number of seat-winning parties of about 2.62 if we assume it is a simple PR system. Instead it is almost always less than 2.0. Of course, STV is not a simple PR system (which would imply party lists). In the case of Ireland, over time, the effective number of seat-winning parties has tended to be almost exactly what the seat product model predicts–as if it were simple PR. It is not possible to say whether Malta’s party system remains less fragmented than expected from the SPM prediction because of its system being non-simple (even more since 1987, as discussed near the end of the post’s main text) or due to features of Maltese politics other than the electoral system. If the latter, it might mean that even nationwide PR would not open up the system as much as expected. Yet I suspect it would do so, as some of the features of Maltese politics that hold down the number of parties–like the clientelism–are enabled by STV. A list system, on its own, should lower the barriers to entry by making “programmatic” vote-earning strategies more viable for competition against the established clientelist parties, even if the seat product were not also increased. Of course, if they change to a nationwide PR system in a 65-seat assembly, the seat product would be 4,225, which would lead us to expect the effective number of seat-winning parties to be about 4.0 (ignoring the impact of a legal threshold). If that is too much, then they should seriously consider districted list PR. (Perhaps with about nine districts instead of the current thirteen; see also discussion of these points in the comments section.) ↩︎
5. Hirczy de Miño and Lane note that only about 1% of votes for the major parties were transferred to a candidate of the rival party. Wolfgang P. Hirczy de Miño & John C. Lane (1996) STV in Malta: Some surprises, Representation, 34:1, 21-28, DOI: 10.1080/00344899608522982 ↩︎
6. See Electoral System Change in Western Europe. ↩︎

Is AV (RCV) consistent with a plurality requirement?

From the Attorney General of Connecticut, a very detailed, thoughtful, and interesting memo on the question of whether the alternative vote (or “instant runoff”–a single-seat version of a ranked-choice system) is consistent with the state constitution’s requirement that elections for state offices be by “the greatest number of votes.” The AG indicates that the question is a “close call” but comes down on the side of RCV being in violation.

He compares rulings and constitutional provisions in other states, including how the state supreme courts of Alaska and Maine reached opposite conclusions on essentially the same question.

I’ve always seen a majority as by definition also a plurality (the reverse is not true!). But maybe a “plurality provision” really can prohibit use of the alternative vote/RCV. It is indeed a close call, and as the memo notes, much of this really does come down to the question of what “votes” means–the initial expression or the full tabulation.

There is also a useful article about the matter in Connecticut News Junkie. Thanks to Jack Santucci (via Twitter) for the links.

What role for “fusion voting”? Limitations and a potential “open” improvement

It was not something I ever would have predicted, but so-called fusion voting is making a comeback, at least as an option that some American electoral reformers are advocating or suggesting should be considered. In the form that advocates generally prefer, “fusion” allows a candidate to be endorsed by two or more political parties, with each of the endorsing parties having its own ballot line or other way in which voters can indicate which endorsing party they favor. This is often called “disaggregated” fusion, because the contribution from each party to the candidate’s total is transparent. It is thus distinct from an “aggregated” version in which a candidate may be formally endorsed by multiple parties but there is no way to differentiate which party a given vote may have come from.¹ Whenever I refer to “fusion” in the rest of this entry, I will mean the “disaggregated” version unless I indicate otherwise. [See text at bottom for footnotes; the intra-post links are not working for some reason.]

The purpose of this planting is to express a somewhat nuanced and qualified support for the idea, alongside a healthy dose of skepticism that it can deliver what its advocates claim to want, and–this being Fruits & Votes–some comparative (non-USA) perspectives and a potential modification that draws from the practice of open-list systems. I think my open-list version of fusion overcomes one of the main problems I identify with standard fusion models. Of course, I do not claim it does not have its own problems or that it is the system folks should now rally behind. The purpose here is just to have a “think piece” and stimulate conversation.

It is possible that in some quarters I have been placed into a “pro-fusion” camp (and at the same time an “anti-RCV” camp). I do not see myself that way (with respect to either “fusion” or “RCV”), but I suspect the perception comes from the recent report I wrote, along with Jack Santucci and Michael Latner, for Protect Democracy and APSA on electoral-system reform options for the USA. In that piece, we suggest fusion as an option for offices that, absent even more fundamental institutional reform, must remain single-seat (like Senate and popularly elected executive offices). However, and crucially, we made the case for fusion only alongside the adoption for the US House of open-list PR (and thus multi-seat districts). At least that is how I understand what we wrote, and it is what I personally intended when working on the essay. In other words, not to endorse fusion for the House or for other offices independently, but as part of the adoption of OLPR. You, of course, can read the report yourself and see if you think we suggest something different from what I am saying here. (We also do indeed give some reasons why STV–which uses RCV ballots–is not the best choice for the US House, in our assessment.)

The logic of fusion for single-seat contests alongside OLPR where practical is that both do–or are capable of doing–the same thing, which is pooling votes across candidates or parties that present a joint slate for election. This is the limited way in which I think so-called fusion voting is useful. Like any form of election via lists, it pools votes in the process of determining winners. In multi-seat districts with proportional representation, it obviously allows any given list potentially to win more than one seat, but regardless of how many it wins, it lets a group of candidates share votes because, prior to the election, a list was structured over which votes pool. In the case of OLPR, this means votes for any of the individual candidates on the list are summed up to determine a list total before the application of the PR method to determine how many seats the list as a whole wins in the district. Similarly, disaggregated fusion means that votes are cast for distinct parties (or for typically for the same candidate appearing on different ballot lines for each endorsing party) and are then summed up, that is pooled, to determine which list of parties endorsing a common candidate has the plurality in the single-seat district.

Based on how the winner of a contest is determined, “disaggregated fusion” belongs to the wider family of pooling systems, and I would much rather that its name be something that reflected this aspect of the rule, rather than highlighting the “fusion” aspect of the vote itself. In my own published work as well as from time to time on this blog, I use the term “fusion” (or “fused vote”) to refer to cases in which a vote for president and a vote for assembly–or any other set of two or more distinct offices–are cast through the same voter choice. In a sense, the use of the term, fusion, in the US refers to something quite different: a vote for Party A is also, in effect, a vote for Party B if they have nominated the same candidate on different lines of the ballot for the same office. The point is that in plain electoral-system terminology, this is pooling not fusion! Nonetheless, the name, fusion, is well established in the US (albeit probably nowhere else), and so I will go on using it, despite my reservations and preference for some alternative name that invokes the rule’s process of determining an election winner.²

Ultimately, what makes me skeptical about so-called fusion voting is precisely that only a single candidate can be on the ballot despite the vote being pooled across a list of parties. We might even say it is normal pooling electoral rules turned on their head. In OLPR, votes for different candidates pool to one party or alliance list. In American fusion voting, votes for different parties pool to a single candidate. It is pooling either way, but the direction of vote flows is reversed. And therein lies a critical problem, if one of our objectives is to foster multipartism without PR, in the service of some wider goals for improving democracy (and assuming PR is not a viable option, at least at present). 

To put it simply, fostering new parties is hard when those parties are inhibited from having their own candidates on the ballot and appealing for votes. Disaggregated fusion lets the parties get their names on the ballot, but the actual process of electing a winner for a given office remains candidate-centered, typically around the candidate of a major party. The small parties are reduced to being “vote funnels” for one of the major parties, but have little opportunity to advance themselves as alternatives. Or they can do so only to the extent that they retain “blackmail” potential whereby they withhold their “fusion” endorsement from whichever major party their votes are closer to, threatening to play spoiler.

It is plausible that the spoiler threat is more credible–thereby inducing policy concessions from a major party–when a small party has a ballot line and can decide whether or not to play the fusion game, contest by contest. In that sense fusion may be a step forward in terms of leverage for minor parties compared to the status quo where fusing (aligning and then pooling votes) is not an option. (The only option is to stand down entirely.) But it is more credible only in the sense that building a multiparty system in the absence of PR requires being willing to be a spoiler at times. You don’t build a new party by being afraid of spoiling, you build it by giving voters a reason to prefer you to the existing major parties and by recruiting candidates to run under your label. A big problem with fusion is it short-circuits the candidate-recruitment side of party building inasmuch as the usual role of the smaller party in a fusion system will be simply to funnel its votes towards a major party candidate that they are unlikely to have had any role in selecting.³ 

The biggest advantage of fusion, in its disaggregated form, and one that advocates often bring up is information. It is said to make clear to the winner, and the public, how much of a candidate’s support came from one party’s voters and how much came from another. In this way, it serves a purpose similar to ranked-choice voting where preference flows can be used to determine the electoral coalition that brought the winner to office. (Again, fusion does this via pooling votes; by contrast, the various forms of RCV do this through “transferable votes.”) A recent Protect Democracy report advocating fusion voting for presidential elections demonstrates the point by showing cases of state-level votes for president over the years in which the winner of the state’s electoral votes needed the votes cast on a minor-party line in order to defeat the other major-party competitor. For example, in New York in 1944, the Republican ballot line had more votes than the Democratic line, 47%–39%, but the latter party’s nominee (FDR) won because the votes cast on the American Labor line (8%) and for the Liberal Party (5%) pooled to FDR to give him the electoral votes. Also worth calling attention to here is Reagan beating Carter for the 1980 electoral votes of New York, despite Democratic votes (44%) being greater than Republican (43%). Reagan also benefitted from pooled Conservative votes (4%), while the Liberal Party (8%) chose the blackmail/spoiler route instead by endorsing John Anderson rather than Carter. 

So fusion can work to demonstrate vote flows and also leave a small party the option of not pooling its votes with a major party. But is it worth the trouble? Maybe. As I noted, it may generate some incentives to form new parties that would not otherwise be present, and once the parties are in the system they may be willing to withhold endorsement (fusion) now and then and advance on their own to the general election, like the NY Liberals in 1980. But it is a far cry from any system in which new parties are encouraged to advance their own candidates and have them on the general-election ballot where they can effectively build the brand and make a case for the alternative they represent yet without being spoilers. 

I will not go further into the cases in the USA where fusion is currently used and what impacts it has had. There are other authors to whom the interested reader can turn for that. I will note, however, that while New York has had disaggregated fusion for a long time, and does indeed have other party labels with current or past ballot lines (e.g., Working Families, Right to Life, Conservative, Liberal), I certainly do not think of New York as standing out as having a flourishing multiparty system on account of fusion. These minor parties play some role in the political process, but it is mostly as vote funnels for Democrats or Republicans. So the notion that fusion, on its own, will foster a multiparty system strikes me as wishful thinking. The notion that it may be a path to adoption of proportional representation seems even more fantastical. Does NY even have a PR movement that is notably stronger than in other states? If so, it certainly can’t claim to have been successful at it.

In spite of this, a recent spate of articles has discussed fusion and advocated for it either forthrightly or conditionally. Lee Drutman had a piece for New America in 2022; the main point is that fusion is beneficial in its own right, but on p. 21 Drutman also advances the notion that its adoption might bring about a “moderate” party whose presence in the system could be “leveraged to support broader changes in the political system that would break the “two-party doom loop” and end the zero-sum nature of American partisan competition, such as proportional representation through multi-member districts.” As I noted already, I have serious doubts that fusion is in any way a path to PR (even if I am on record saying it could be worth adopting for single-winner races along with PR for the House). Drutman’s case for fusion is mostly that it would be a way to promote a centrist party that could fuse (pool) with, in principle, whichever of the major parties had the more “moderate” nominee. Steven Hill offered a critique of Drutman’s piece, in which one point that resonates with my argument here is that the New York Green Party rejects fusion because it sees it as “cover” for the big parties in a “horse trading” game among political machines rather than a way to build an alternative brand. Then Drutman had a further response in which he argued fusion was the more “party-centered” option than RCV. (I noted above that I see fusion as remaining fundamentally candidate-centered.⁴) Then Hill had a further response, the main point of which is captured by the subtitle: “Fusion yes, but more study of modern-day uses needed,” specifically noting that many of the claimed benefits of fusion have not been not clearly substantiated. He reiterates his view that “RCV” is better for building a multiparty system. (I am skeptical of the latter claim, but that’s beside the point for this planting.) I am not going to attempt to summarize all the points these two authors make, but anyone interested in understanding the debate over fusion should read them all. 

More recently, Protect Democracy has come out with the report I already mentioned above, suggesting the adoption of fusion for presidential elections–at the state level for electors. It is a thoughtful piece and I recommend it, but ultimately it encapsulates most of my reasons for being skeptical about fusion. A central thesis is captured by this quote from the introduction: “By empowering factions with differing views on policy but a shared commitment to liberal democracy to unify in support of a single candidate, fusion can serve as a key tool for defeating authoritarian threats at the ballot box.” The introduction specifically mentions possible third-party or independent candidates for 2024 like Cornel West or a nominee of “No Labels” (which is, of course, a label). My skepticism lies in whether potential third-party presidential candidates are committed to liberal democracy as their main animating purpose, and would be content to be vote funnels for Democrat Joe Biden in order to stop Donald Trump, or whether their main purpose is to make a point, try to get into debates, and advertise themselves. Maybe they, or key supporters, actually want to be spoilers. And then there’s the question of whether Biden would even want to have his name listed along with a party that featured Cornel West as its most prominent national figure. (Fusion rules permit the big party, or its nominee, to reject the candidate’s name being on a ballot line with which they do not want to be associated.⁵) Fusion seems like a potential dead end to me, at least so long as the minor parties do not exist with avenues for development and influence through actually electing their own people to office and participating in legislative coalitions–something that PR for the House (and state legislatures) would facilitate, of course.

To what extent does disaggregated fusion exist anywhere else? Is it just an “exceptional” American curiosity? As I said above, if we think about it as part of a wider family of pooling electoral rules, obviously pooling exists anywhere that has list PR. But I mean more narrowly, as a means of facilitating pre-election alliances between parties that jointly nominate the same candidate for some office. Of course, pre-electoral coalitions are not unheard of in single-seat contests around the world. India has them as an established feature of its politics, for example. However, they are not “fusion” because only the party name/symbol of the candidate being nominated by such an alliance in a given district appears on the ballot. Other parties in the alliance merely commit to mobilizing their supporters behind the candidate (and party) that has been granted the alliance’s nomination in the district. The alliances are public, but not on the ballot, and even if they were, voters would be voting for the “aggregated” alliance, and not selecting a party below the level of the alliance.

The cases I know of that may qualify as disaggregated fusion can be found in Latin America. For instance, Venezuelan presidential ballots in the 1958-88 period used to permit multiple parties to endorse the same presidential candidate but retain their separate ballot symbol. The parties endorsing a given presidential candidate were not grouped together, but the picture and name of the presidential candidate were included along with the party symbol on the square on the ballot that the voter would mark.⁶ I believe Mexican Chamber of Deputies elections at times have used something that could count as US-style fusion. Mexico has a mixed-member majoritarian (MMM) system, and parties can run separately in the list-PR component while endorsing a common candidate for a single-seat district. I recall that in some years voters could place their vote on one of the different party symbols backing a given candidate. [Details–click the link for some cool ballot images!] It is also possible that Italy has had this option in some of its elections. I hope someone can confirm or correct this.

In the end fusion as understood in the USA is rare, but not unheard of. It has some potential for allowing multiple parties to form that might not otherwise under current rules, but those parties will tend mainly to be vote funnels for a major party. The number one defect from a multipartism-fostering perspective is that the parties that are pooling votes (by “fusing” on a single candidate) do not each get their own candidate on the ballot. This seriously hampers party formation and development, especially in an electoral process that remains fundamentally candidate-centered despite the use of fusion. A final question is whether there may be a way to improve so-called fusion voting to overcome this defect. 

The most obvious improvement I can think of is to take the lead from other pooling systems and adopt open-list plurality. Call it “open fusion.” Like disaggregated fusion, the separate parties would have ballot lines, but unlike currently existing US-style fusion these parties could also have their own candidates. But by being grouped on the ballot together, it would be clear that a vote for any of them is a vote for the group. The winner is the plurality candidate within the plurality group. This is perhaps its potential biggest downside politically, in that the winner need not be the individual candidate with the most votes; this is of course true of single-winner RCV as well, when considering first-preference votes. (Two-round versions are also imaginable as a solution to this, but add complexity and cost via another voting day.⁷) Open list in single-seat districts is in use in one district in Finland and one in Peru. These countries both have open-list PR systems and thus use multi-seat districts, but each also has one district that elects only one member, while permitting lists to have more than one candidate.⁸ 

Of course, my “open fusion” model has flaws. There does not exist an electoral system with no flaws. One flaw that immediately comes to mind is the one I alluded to above: what if one of the candidates does not want any association with other parties (and now also candidates) that might like to direct votes towards him or her? Of course, no candidate would be forced to accept being part of any such group. The downside would be that no one plays the game of open fusion because no one wants to be associated with the (perceived) extremist parties that would funnel votes to a given major-party nominee. But this is already an issue with disaggregated (but “closed”) fusion, as noted earlier. Is it worse when the vote-funneling minor parties have their own candidates on the ballot instead of just a ballot slot? Maybe, but also maybe not. Perhaps parties emerge under open fusion that are more serious inasmuch as the proposed system provides incentives for pre-electoral cooperation without depriving smaller parties of the ability to show who they are–literally, by nominating a candidate and making their own case. One could imagine ways to design ballots so that voters could reject a candidate in a group (meaning their vote could not pool to the candidate so-rejected). However, this risks making such a ballot more complex and error-prone, which is precisely one of the arguments against ranked-choice ballots (see Cormack 2023 and Pettigrew and Radley 2023).

I know many readers and reform activists will say, just adopt the alternative vote (a common “RCV” system). I am not going to entertain that option here, in a planting that has already grown rather large. Suffice it to say that there are plenty of reasons to doubt “RCV” is the answer, and even if it is by some objective standard, it faces obstacles to adoption and retention. If RCV has serious problems–which not everyone agrees it does–then maybe that leaves fusion, of some sort, as long as we are stuck with single-seat district elections.⁹ As I have argued here, fusion as practiced in the few jurisdictions of the US that have used it, is not very promising as a means to foster a genuine multiparty system if adopted on its own. It has some redeeming qualities, however, and one its biggest flaws–reducing minor parties to being mostly vote funnels without their own standard-bearers–could be fixed by having an open-list system in the single-seat elections.

Notes

Something has gone fundamentally haywire today with the Word Press footnote function. I was able to save them all as text, so here they are, and I think the footnote markers above are all in the correct place with numbers. But the text for all but one of them vanished from the online editor between draft and saving/posting and an attempt to paste the text back in at the footnote markers again failed. And then for some reason, footnote 6 reappears below the following block of text as “footnote 1” even though it does not appear in my editor (or else I would edit it out). This is bizarre and frustrating. I hope WP fixes this. In the meantime, I will be back to manually placing footnotes.

1. California has this aggregated fusion. It is possible for two parties (e.g., Peace & Freedom and Green in the most recent election) to endorse the same presidential candidate. I believe both party names are then included beneath the candidate’s name, which advertises the candidate as a “fusion” candidate. However, there is only one line and one voting bubble to mark if one wants to vote for this candidate, which is what makes it “aggregated.” Oregon and Vermont have similar provisions. Apparently there is (or was, as of 2013) a possibility of aggregated fusion in the UK as well, and is an option in South Australia (where it is, of course, combined with the alternative vote, a “ranked choice” system). 
2. This is related to my general disdain for calling electoral systems “voting systems” as if voting were all they were about, as opposed to allocating political power for which voting is only one critical part of the process. 
3. They especially will not have such a role when candidate selection is via primaries. Perhaps they could attain such a role through incorporating multi-party fusion somehow in primaries. I am not sure I like this idea, let alone have any idea how it would work. So I will just leave it here as something to ponder. 
4. Whether it is more or less so than RCV is a separate question and one I will not entertain here. 
5. Drutman (2022, pp. 19–20) mentions this option of a major party refusing fusion as a reason why the rule would be unlikely to foster genuinely extreme parties (an advantage that RCV does not have, because the major party candidate is not in a position to reject individual voters ranking him or her after casting a higher preference for the extremist). 
6. These Venezuelan ballots also featured “fusion” in the sense that I have used the term in various research publications and sometimes on this blog: For each party there was both a “large card” and “small card” (all on the same ballot paper despite the term, card). The large card was for the presidential vote, as as noted fusion nominations were permitted whereby votes from different parties endorsing the same candidate would pool to determine the winner. The small card was for a series of other offices–Senate, Deputies, and state assemblies (as well as local councils before those started to be elected on a separate date)–hence being a “fused” vote in the sense that one vote counted for several distinct contests. A voter could split the ticket, but only between president and everything else, and the voter could vote for a given presidential candidate without voting for the president’s core party. 
7. For instance, if the candidate with the most votes is not in the alliance with the most votes, have a runoff between the top-voted individual and the leader of the largest alliance. I am not sure I like this, and there may be better ways to deal with it aside from simply accepting a non-frontrunner as winner. (Or such a runoff could be triggered only if the plurality alliance was short of 50% or some other (lower) threshold and/or if the leading candidate overall whose alliance total was second had not cleared some individual vote-percentage threshold. I don’t really like such complexity, but I could see some political benefit to making a second round required under some limited circumstances.)
8. Uruguay previously elected its president via such a rule. The vote was also fused in the comparative electoral systems terminology sense that I noted above: a single vote counted for a presidential candidate, the alliance of which he or she was a part, and a list of candidates for each of the two houses of the national legislature. (Especially in earlier years, the lists were actually factional, within a major party, so “alliance” can also refer to the larger party rather than a multi-party grouping. In addition, the factional lists for congress also pooled to the larger alliance/party.)  
9. Perhaps one could imagine ways of combining (disaggregated) fusion and RCV. However, it might be too complex and I will leave it as an exercise for the interested reader. (In the last piece of the multiple-essay exchange between Hill and Drutman, linked above, Hill suggests an interest in “innovative hybrids” of RCV and fusion but does not elaborate.)  

1. These Venezuelan ballots also featured “fusion” in the sense that I have used the term in various research publications and sometimes on this blog: For each party there was both a “large card” and “small card” (all on the same ballot paper despite the term, card). The large card was for the presidential vote, as as noted fusion nominations were permitted whereby votes from different parties endorsing the same candidate would pool to determine the winner. The small card was for a series of other offices–Senate, Deputies, and state assemblies (as well as local councils before those started to be elected on a separate date)–hence being a “fused” vote in the sense that one vote counted for several distinct contests. A voter could split the ticket, but only between president and everything else, and the voter could vote for a given presidential candidate without voting for the president’s core party. ↩︎

The Italian Government seeks a new constitutional reform

The following is a guest planting about the Italian government’s constitutional reform proposal by Gianluca Passarelli, and it addresses various questions that we had in previous discussions of this topic (1, 2).

After the attempt made in 2016 when the parliament approved a reform promoted by Prime Minister Matteo Renzi (Democratic Party, centre-left) but that was rejected via a constitutional referendum, Italy seems to be dealing again with a shot to deeply modify its institutions. The reform is called by the new Prime Minister Giorgia Meloni (Brothers of Italy, right-wing, first woman to cover that position) leading a neo-fascist party in a coalition with the League (far right regionalist party) and Forza Italia (Silvio Berlusconi’s heritage): the project aims to introduce the direct election of the head of the government. 

After weeks of internal discussion and considering the political hostility and the citizens’ perplexity (it’s a divisive argument) for the introduction of a presidential regime, the right coalition moved towards a different option. However, instead of generating more clarity and a softer model, the proposal goes in the direction of assuring a direct election of the head of the government. 

The document issued by the minister Elisabetta Casellati indicates that the head of the government will be “directly elected” by voters. Basically, in a way, the proposal resounds and mimics the Israeli experience of the model used between 1996-2001 after its adoption in 1992. But the similarities end here, at the “direct election”.

The electoral system for the Parliament (Italy has a bicameral system: Senate, 200 members, and House, 400 members) indicates so far a PR with bonus, a majority-assuring system that would confer 55% of the seats to the list(s) linked to the candidate to the premiership. But there is not a clear model, nor the indication of whether or not preferences would be allowed, which formula will be adopted, in how many districts Italy will be divided into. 

According to the recently circulated draft, the project of constitutional reform would aim to modify three articles of the Italian Constitution, as follows. The presidential prerogative to appoint the PM (art. 92 Cost.) would change so that the President has to indicate the name of the candidate winning the elections. As head of the state, the President would keep the prerogative to appoint the ministers, albeit based on the premier’s indications. Moreover, one main hypothesis circulating is that were the PM to end his/her term before the legislature, the President should indicate a new head of the government by choosing among those elected in the list(s) of the candidate who won the elections. So, there is not the provision of a constructive vote of no confidence (as in an earlier version–ed.). And, due to the previous proposed changes, the reform would modify the norm regulating the vote of confidence and the motion of no confidence to the Government (art. 94 Cost.). Consequently, the presidential power to dissolve the parliament (art. 88 Cost.) is under discussion due to the provision that indicates the President has to appoint the head of the coalition who won the election and to appoint another politician from his/her majority in case s/he resigns. The President would become a sort of notary taking notes of political changes but without intervening. The proposal also would cancel the presidential power to appoint life senators, whilst those still in office would keep their seats as well as the past presidents of the republic that would become life senators.

 As the proposal would move to the parliament for a political debate, we will probably know more about this new project of constitutional reform. In particular, the absence in the proposal of a minimum threshold to win the election and to obtain the bonus would likely pave the way to a pronunciation of the Constitutional Court. In fact, in 2015 Judges have already declared that a reasonable threshold should be included in the electoral reform (at the time it was 40%). 

Finally, the text is a prelude to a new referendum in a few years because it is very divisive. However, we will see if the proposal will be modified to move towards a shared (two-thirds of the parliament) reform that will exclude a referendum.

Netherlands 2023

Update to the below: Exit poll says Results show the PVV (Wilders) on 35 37 seats. That would be a much bigger lead than pre-election polls suggested. However, it is only 24.7%, and not exactly the “landslide” that several news headlines are claiming. The Green Left/Labour alliance is second (25) and VVD third (24). I will not speculate on possible government options, and will leave the rest of this unchanged.

The Netherlands has a general election today–an early one, the previous one having been March, 2021. I have not been following so I can’t say much of substance, but I note from the polling that three parties have been running close and sometimes trading narrow leads in recent polling, but neither has even broken 30 seats in any of the polls going back several months, and collected at Wikipedia. That is out of 150 seats, so we could be looking at a largest party holding fewer than 20% of the seats.

A largest seat-winning party at no more than 20% is something that happens only very rarely in established parliamentary democracies. In fact, in the dataset I use for my electoral systems/party systems analysis, I find only five elections in which the largest party had 20% or fewer: four in Belgium and one in Latvia (1995, with 18%). That is out of 293 elections in such systems. Moreover, the Belgian party system is one where the Flemish and French speaking electorates have their own parties within larger party families. (I don’t know much of anything about Latvian politics.)

So, while not unprecedented, the outcome the polls are suggesting we could see would be extremely rare. The three parties vying for the top spot are (1) the VVD, a center-right party that has led several recent governments under outgoing PM Mark Rutte; (2) GL-PvdA, an alliance of the Green Left and the Labour Party; and (3) the PVV, the anti-immigrant party led by Geert Wilders. Rutte is no longer leading the VVD, having been replaced by Dilan Yeşilgöz-Zegerius. The GL-PvdA leader is Frans Timmermans, a former foreign minister who also has held several EU positions over the years, including posts focused on climate policies.

Also, it should be obvious, but it perhaps is not, and I even saw a political scientist (who is not a specialist on Dutch politics) get this wrong on social media: Coming in first, if it happens, will not put Wilders on the verge of becoming Prime Minister. Negotiations on forming a government will take a while, and the leading party might not even be in the coalition that eventually is formed. However, I do not know what coalition combinations have been put forth during this campaign. I hope someone reading knows, and can enlighten us in the comments.

The Netherlands uses nationwide proportional representation (flexible list) in a 150-seat district. There is no legal threshold beyond the requirement that a list earn a simple quota (1/150) of the nationwide votes to qualify for a seat. This means the seat product is 22,500, from which the seat product model would lead us to expect a largest list seat percentage of 28.6% (around 43 seats). Over time, the country has basically nailed this expectation, with a mean of 29.3%. The previous minimum was 20.7% in 2010; in the 2021 election it was 22.7% when the VVD won 34 seats. The last time the leading list won 30% was 1998. Thus the 2023 result is expected to be a continuation, indeed an acceleration, of a recent trend.

Taiwan opposition parties forming alliance

As the deadline for submitting candidacies approaches, the two main opposition parties in Taiwan have announced they will submit a joint candidate. They have yet to determine which party’s candidate will stand down in favor of the other– Hou Yu-ih of the Nationalist Party(KMT) and Ko Wen-je of the Taiwan People’s Party (TPP). They have a plan to use public opinion polls to settle the question, with each side selecting pollsters whose polling is to be considered. The election will be on 13 January 2024.

The presidency is elected by direct plurality, and up to now the incumbent Democratic Progressive Party’s candidate has been leading polls consistently, but recently with only around 40%. Hou and Ko have been roughly tied recently, in the mid-twenties, with polls varying as to which of the two is ahead of the other. If they could successfully combine their votes, the alliance should win. But alliances are rarely so seamless.

In addition, the KMT and TPP promise to form a coalition cabinet based on the parties’ relative success in the concurrent assembly election. Here is where things get really tricky. Will they make the decision according to seats won? This is the basis on which coalitions normally parcel out posts, after all. But Taiwan uses a disproportional system–specifically, a mixed-member majoritarian (MMM) system. It is unclear to me whether the parties are also forging an alliance for the assembly election. [See update below.] Under MMM, and with a joint presidential ticket, it would make sense to do so, given that this type of mixed-member system makes the single-seat district (plurality) contests more important to the overall seat balance. (They still could run separately on the list, given the two-vote system.)

In any case, the balance of forces in the legislative assembly will not necessarily reflect the balance of voter preferences, either due to vote-splitting (if the KMT and TPP compete against each other in single-seat districts) or “contamination” effects (whereby if one partner has the local candidate and the other does not, the one with the candidate may draw additional list votes). Votes for parties in the nominal and list component often have been fairly divergent in past Taiwanese elections under MMM. For instance, in the 2020 election, the DPP won 45.1% of the nominal (constituency) vote and 40.0% of the party-list vote. For the KMT the figures were 40.7% and 33.4%.

While pre-electoral coalitions are hardly unheard of in disproportional systems (see India under FPTP, plus for an MMM case, Japan), they may be more difficult when they involve a concurrent presidential election, given the importance of that indivisible office. They may also be more challenging to hold together when the allying parties’ strengths are quite balanced, as opposed to there being a clearly dominant partner (like the nationwide parties in the alliances in India or the LDP in Japan). I am referring to their balance in the presidential polling. It is not clear to me whether there is polling for assembly voting preferences. In the current assembly, the parties are very unbalanced. In 2020, the KMT won 38 seats and the TPP 5 (with DPP winning 61). In the nominal tier (single-seat districts), it was, not surprisingly, even more unbalanced: KMT 25, TPP 0 (DPP 48) [Corrected].

I should note–this being Fruits and Votes–that Taiwan uses a semi-presidential system. However, it is of the president-parliamentary subtype, the one in which the president retains constitutional authority to dismiss a premier and cabinet. Thus holding together pre-electoral coalitions after an election and throughout a term would tend to be more difficult than under a parliamentary (or premier-presidential) system.

The alliance seems quite fraught, especially as it is being struck so late. And, while he does not address the assembly elections, Nathan Batto of Academia Sinica is referenced in the Economist article saying that some of the TPP’s “Gen-Z” supporters may object to teaming up with the conservative KMT. Indeed, alliances are not necessarily as large as the sum of their parts. If the parties can hold each others’ support in the alliance, they should be able to defeat the DPP. But that is a big IF.

At Frozen Garlic, Batto has interesting detail on the question of how to use to polling to sort out the alliance presidential candidate, and states that Ko (TPP) got “rolled” in the process they agreed to.

Update: I asked Batto my question about the legislative elections, and he responded at his blog. The TPP is such a small party that it does not have many single-seat district nominees. The “overwhelming majority” of contests will be KMT vs. DPP, but there will be a few districts where the KMT and TPP both have candidates. So (my conclusion, not Batto’s) the TPP is a highly presidentialized party in that it potentially can win a significant share of the votes (per polling) for its presidential candidate, and perhaps also for its party list (especially if it runs its own candidate for the presidency to generate some coattails), but it has little presence as an organized party otherwise.

Later update: This plan did not go well.

2023 Swiss election: minor technical problem leads to major embarrassment

Like most European countries, Switzerland uses proportional representation (PR) to choose members of the National Council, the lower house of the country’s federal legislature. However, the Swiss system has distinctive characteristics which require a special procedure to determine the nationwide party percentage shares, and a minor technical problem during the federal election held last Sunday, October 22 resulted in the calculation of erroneous nationwide party percentages figures, which remained undetected for nearly two days. Although the Swiss government issued corrected figures on Wednesday, October 25, the unfortunate incident has become a major embarrassment for a country which prides itself on precision. In this posting we review the procedure used in Switzerland to determine nationwide party percentages in federal elections, as well as the technical issue that generated the erroneous figures and the steps that could or ought to have been taken to prevent its occurrence.

In elections to the 200-seat Swiss National Council voters in twenty multi-seat cantons have as many votes as there are seats to be filled; in 2023 that figure ranged from a minimum of two to a maximum of thirty-six. In these cantons, voters can select a single list, thus voting for every candidate on it, but they can also drop a candidate from the list, and either put another candidate from the same list a second time and cast an additional vote for that candidate (a procedure known as cumulation), or write in the name of a candidate from another list (a practice known in French as panachage). Moreover, voters can also compose their own lists by combining candidates from different lists.

Meanwhile, the remaining six cantons have one seat each, filled by plurality voting, with voters choosing a single candidate. Since adding up the canton-level results to obtain nationwide totals for each party would result in skewed figures, due to the disproportionate weight of the vote totals in the larger cantons, it is necessary to convert the results from the multi-seat cantons to comparable values on a nationwide basis.

One alternative would be to divide the votes obtained by each party in a multi-seat canton by the corresponding number of National Council seats to be filled. However, due to the incidence of under-voting – some Swiss voters choose fewer candidates than the number of cantonal seats – the sum of comparable values for all parties would be smaller than the number of valid ballots. Instead, in each canton Switzerland’s Federal Statistical Office (FSO) divides the number of votes obtained by each party by the total number of valid votes, and then multiplies it by the number of valid ballots, so that the sum of comparable values for all parties will be equal to the number of valid ballots. Then, the canton-level comparable values for each party are added up to obtain nationwide totals, and the corresponding party percentage shares are calculated by dividing each party’s nationwide comparable value total by the total number of valid ballots.

By the way, FSO calls the comparable values “fictitious voters” – an incredibly unfortunate term in an era of rising skepticism about the transparency of election processes, and all the more so in light of the incident addressed here.

At any rate, the procedure utilized by FSO delivers modestly inflated comparable values at the expense of under-voting (whose incidence in Swiss federal elections is fairly low), and which also vary slightly when calculated at different geographical levels, precisely because of under-voting percentage variations. More importantly, it contributed directly to the erroneous calculations, because it depends on accurate valid ballot totals. As it was, according to FSO an undetected data import glitch triggered by a programming error caused voter turnout data records for the single-seat cantons of Glarus, Appenzell Ausserrhoden and Appenzell Innerrhoden – which deliver their election data in a non-standard format – to show up multiple times, thus inflating turnout statistics for these cantons by a factor of three to five, including the crucial valid ballots statistic. Moreover, the faulty data import routine wasn’t even tested with previous election results before being used on election night.

(Specifically, the affected cantons’ commune-level records in the CSV-format turnout data file were repeated three to five times; the canton-level records showed up once, but with values inflated by the number of times the commune-level records were repeated. However, this problem did not occur with the corresponding records in the party votes CSV data file, where only the parties’ comparable values in the three affected cantons were inflated.)

Normally, the comparable values for parties in single-seat cantons are equal to the number of votes polled by their respective candidates in those cantons, since voters in such cantons can only vote for one candidate, and therefore the total number of valid votes for all candidates is equal to the total number of valid ballots. But in the case of the three affected cantons, the inflated valid ballot totals in turn inflated the comparable values for the parties, thus skewing the nationwide totals for each party, which in turn produced incorrect party percentage shares at that level.

For example, in the canton of Glarus the Swiss People’s Party (SVP) polled 5,388 votes out of 12,650 cast for the four candidates contesting its single seat, and the comparable value for SVP should have been equal to its number of votes, as 5,388 divided by 12,650 votes, and then multiplied by 12,650 valid ballots equals 5,388. But since turnout data for Glarus was repeated five times, the division of 5,388 by 12,650 was multiplied by an inflated total of 63,250 valid ballots, resulting in an inflated comparable value of 26,940 for SVP in the aforementioned canton.

In all, the turnout data errors inflated the nationwide valid ballot total for the October 22 National Council election by 109,715, from 2,554,482 to 2,664,197 (and from 33,287 to 143,002 in the three affected cantons). In particular, the nationwide comparable value totals of SVP/UDC, The Centre (DM/LC) – the 2021 merger of the Christian Democratic Party (CVP/PDC) and the Conservative Democratic Party (BDP/PBD) – the Liberals (FDP/PLR) and the Social Democratic/Socialist Party (SP/PS) were inflated by 47,194, 29,158, 19,119 and 11,840, respectively. In terms of party percentage shares, the figures for SVP/UDC and LC/DM were overestimated by 0.62% and 0.52%, respectively, while those for the Greens and the Green Liberal Party (GLP/PVL) were underestimated by -0.40% and -0.31%, respectively.

Even worse, the erroneous figures remained unchanged for almost two days after the election. According to news reports, in a hastily arranged press conference on Wednesday, October 25, a contrite FSO director apologized for the faux pas, indicating that due to a work overload quality controls that ought to have been carried out at an earlier stage didn’t take place until the afternoon of Tuesday, October 24, at which point the problem was detected and corrected the following day.

A press release [PDF] issued by Switzerland’s Federal Department of Home Affairs has further details, including both the originally published and updated nationwide party percentage shares.

Meanwhile, in the morning of Tuesday, October 24 – early afternoon in Switzerland – I came across the same errors while attempting to replicate the nationwide party percentage shares published by the FSO from the election data files furnished by the agency. Once I realized it was an obvious mistake on their part, I documented my findings on a Twitter thread – which included corrected figures – and notified the FSO about them.

Quite predictably, Swiss news media hasn’t taken kindly to the matter, with one leading German-language newspaper bluntly describing it as a “fiasco,” while wondering if there are further errors waiting to be found; FSO insists there are none, and as far as I can tell that is indeed the case. Another German-language newspaper was even less charitable: writing about the “historic election mishap,” it dubbed FSO as the “Federal Office for calculation errors.”

While canton-level party percentages, and most importantly the distribution of National Council seats and its elected members, were in no way affected by the election data errors, the nationwide party percentage shifts placed the Liberals back in third place, narrowly ahead of The Centre. Before the nationwide party shares were corrected, there was speculation the latter might press for an additional seat at the expense of the former in Switzerland’s ruling seven-seat Federal Council, on account of having apparently finished third both percentage- and seat-wise; Federal Council seats are currently distributed on a 2-2-2-1 basis among SVP/UDC, SP/PS, FDP/PLR and DM/LC.

It’s been speculated as well that the Greens might press for representation in the Federal Council, in light of their improved showing in the updated nationwide results, even though the party’s updated share of the vote still fell by 3.4% compared to the 2019 National Council election.

Although nationwide party percentage shares have no bearing in the distribution of National Council seats, carried out entirely at the cantonal level, they are often perceived as a more accurate indicator of party strength than lower house parliamentary representation, which usually is slightly distorted by the country’s canton-based PR system. For example, in this year’s election Switzerland’s largest party, the right-wing SVP/UDC won 31% of the seats in the National Council (62 of 200) with a nationwide share of 27.9%.

It should also be noted that while press reports have referred to the incident as a vote counting error, a more accurate way to describe it would be as a data aggregation error, which nonetheless underscores the importance of timely verification by human beings of automated processes, all the more so given that the problem at hand could have been detected fairly quickly, and dealt with before it became a national embarrassment. Specifically, the fact that three single-seat cantons had party comparable values several times larger than their respective vote totals (and on top of that by whole integer factors) should have immediately raised a red flag: the ratio of votes to comparable values should be a figure very close to the total number of seats for multi-seat cantons, and exactly one for single-seat cantons. A subsequent turnout data check would have then brought up the inflated statistics – including crucial valid ballot totals that sometimes exceeded their respective cantons’ population figures – and the repeated commune-level records.

Incidentally, running the comparable value calculations under the alternative seat divisor method, which didn’t depend on the faulty turnout data, would have produced party shares nearly identical to the corrected figures; under the seat divisor method the parties’ comparable values in one-seat cantons are also equal to their respective number of votes, since the divisor equals one.

Meanwhile, the head of the Department of Home Affairs has ordered an administrative investigation “to analyse and improve the processes together with the FSO.” In the future, a more wide-scale automated plausibility check of the calculations will be implemented, along with the deployment of more staff to conduct controls on election day, as well as a “complete review of the overall processes and control modalities.”

Time will tell if the FSO blunder in the October 2023 National Council election has left any lasting damage. That said, the sad fact is that incidents like these undermine public confidence in the government’s ability to run election processes in a transparent manner, while potentially fueling election conspiracy theories.

Finally, on a personal note I can say that the world has truly become a global village when an elections specialist in Puerto Rico like myself can spot and diagnose an election results error in faraway Switzerland, and have the accurate figures online a day before the Swiss government finally corrected them.

New Zealand 2023

Saturday 14 October is New Zealand’s general election. Here are a few things I am watching with interest.

There will be a by-election in November, which will increase the size of parliament. In the electorate (constituency) of Port Waikato, one of the candidates died after the deadline for candidate registration passed. The candidate, Neil Christensen, was running for the ACT Party and we can safely say that he had no chance of winning the plurality of the vote in this electorate. Nonetheless, under a strange provision of New Zealand electoral law of which I was not previously aware, the vote in this electorate can not go ahead and so the seat will be held vacant till the by-election, which is set for 25 November. Stranger still, this does not mean there are set to be elected only 119 MPs on Saturday. Rather, the full 120 will be elected, 71 electorates and 49 list seats. (There could be more, due to overhangs, but that is a separate issue to which I will return.) When the by-election is completed, there will be an additional seat in parliament. Voters in Port Waikato will still cast list votes on Saturday, but not electorate votes. And early electorate votes already cast before Christensen’s death will simply be discarded. Graeme Edgele, a lawyer who has made submissions in past electoral-law debates, explains this provision in detail, and other means that he think would be better to deal with this situation, in a recent episode of On the Campaign, a podcast of the NZ Herald.

How many electorate seats will the Greens win? Until recently, the Green Party almost exclusively relied on winning party list seats. Nonetheless, it contested electorates, something that at one time was controversial within the party (per conversations I had with party staff and a then newly elected MP back in 2011). I have written before about why it makes sense for a smaller party like the Greens to have electorate candidates even if they have no serious chance to win them. Things have shifted–the party now has a realistic chance of winning more than one electorate. They are explicitly targeting some urban seats in Auckland and Wellington, as noted in July in the NZ Herald:

There are already an unprecedented four two-tick campaigns for the party ahead of this year’s election in October, including Chloe Swarbrick again in Auckland Central, Julie Anne Genter in Rongotai and Tamatha Paul in Wellington Central.

The reference to “two-tick campaigns” means seeking electorate (nominal) votes as well as party (list) votes. Usually the Green Party goes specifically for list votes, and a high percentage of its voters will vote strategically for the local Labour Party candidate. In these electorates, the Greens are perceived as strong enough that they might win the seats with Labour party voters shifting to the local Green candidate. Swarbrick won her electorate in the 2020 election.

Will Meka Whaitiri win her electorate? Meka Whaitiri was a sitting minister in the Labour government who defected and sat as an independent for the rest of the outgoing parliament’s term. She is now contesting lkaroa-Rāwhiti, one of the Māori electorates. This previously has been a safe Labour electorate. If she beat the Labour candidate, it would not change the number of seats Labour wins overall, due to the compensatory nature of MMP. However, it could help generate an overhang, whereby parliament would expand to 121 or more (even before the aforementioned by-election) due to Te Pāti Māori winning more nominal (electorate) seats than its list vote total would entitle it to proportionally. Some local conversation has centered around whether Māori voters should vote for the Labour candidate on the grounds that Whaitiri may get in regardless, being ranked third on the Te Pāti Māori list. This would be good enough only if TPM gets the 2.5–3% of nationwide party votes that some polls suggest it will, not the under 2% that others show–in other words, if there are not overhangs generated by currently unexpected wins in other Māori electorates. On the other hand, the Labour candidate inlkaroa-Rāwhiti, Cushla Tangaere-Manuel, does not have a party-list nomination. The conversation is whether by voting for the local Labour candidate but giving TPM the region’s list vote they could effectively get two regional MPs. Thus there is a good “effects of MMP” angle here, too! This is discussed in the same podcast episode I liked to earlier. That podcast also discusses how the two major-party candidates in the Napier electorate–the general electorate that is located partly within Ikaroa-Rāwhiti–are campaigning on a promise to widen an expressway through the Hawkes Bay region and building a hospital in the area, as well as emphasizing the need for more more recovery assistance from a cyclone that hit the region. This is a clear reminder that MMP does facilitate the raising of local issues even in a context of overall nationwide proportionality among parties outcome (did someone say “best of both worlds”?).

Will TOP get in? TOP is The Opportunities Party. Most recent polls have it in the 2–3% range, evidently safely below the threshold of party-list votes. However, New Zealand’s threshold is 5% of the list votes or a single electorate plurality. It may be a long shot, but there is at least some chance TOP could win an electorate, Ilam in Christchurch. Its candidate there is former city council member Raf Manji, who contested this seat as an independent in 2017 and came second, with 23.4% of the vote, about half what the victorious National Party MP won. That MP is Gerry Brownlee, a senior figure in National, but he is not contesting again. It would be a stretch for Manji to win, but perhaps some local Labour voters would vote strategically for him. If he were to win, that would qualify TOP for list seats as well, and on current polling it maybe would end up with a caucus of three. TOP is a centrist party, emphasizing investing in youth, and not surprisingly, its website front page includes a clear explainer about how it could get multiple seats even if it does not reach 5% of the party vote. In an additional reminder of the importance of local issues under MMP, Manji also emphasizes the ongoing recovery needs of the Christchurch area due to the destructive earthquake of 2011 among other local issues, as well as implications of the lessons of the earthquake for other regions.*

Beyond election day itself and looking ahead to government formation, a huge question is whether a coalition of National, ACT, and NZ First could form and if it did, could leaders David Seymour (ACT) and Winston Peters (NZF) avoid being at constant war with one another? (Debates and sniping in the media are not promising on that front; and sparring between them is not exactly a new phenomenon.) Will ACT make good on its threats to insist on a “confidence only” deal with National if certain ACT policy priorities (some of which National has explicitly rejected) are not advanced, instead of either a full coalition or a confidence-supply deal? This would mean negotiating spending policy on a case-by-case basis and would risk a level of government instability that New Zealand has not experienced since the first MMP-elected parliament (1996-99). National has even warned of the possibility of no deal (specifically in this case with NZF) and a second election, in a rather desperate-sounding appeal to consolidate center-right votes around National.

Through much of the campaign a National-ACT majority coalition looked likely, which may be what prompted Seymour to make his bold policy demands. In turn, that may have prompted some voters to shift to NZF as a least-bad option to limit ACT influence. It was not long ago that polls cast doubt on whether NZF would reach 5% (and it has no realistic chance of any electorate win). It is by now no longer a given that the center-left is going to be unable to form a new government. Labour (which won a single-party majority in 2020) with support of Greens and NZF might be possible, depending on the precise balance of the results; this potentially also would require involvement of Te Pāti Māori. Maybe TOP would be involved in a confidence-and-supply deal, if it even makes it into parliament. (Potentially it seems it could work with either National or Labour.) It is worth recalling that the 2017-2020 government was Labour+Greens+NZF, although it was not always the smoothest partnership. This is looking like it could get very interesting!

________

* In the past, I would have voted Green if I had a vote in New Zealand. I counted that party among my personal favorite political parties worldwide. However, I would probably vote TOP if I had a vote this year (even if I did not live in Ilam). For one, I rather like a lot of their policies and their “teal” positioning–similar to the set of independents who won seats in the last Australian election. More to the point, I could not possibly vote Green this time, given statements made this week by a Green MP (who is ranked 8th on the list and thus safe) and even shared on social media by one of the co-leaders.

Spain 2023

Voters in Spain have voted in a snap general election today. Polls suggest the support for both of the large national parties, PSOE on the left and PP on the right, has been growing late in the campaign. However, no one expects either major party to be close to a majority on its own, as the PP is leading on only around 33–35% the vote per final polls. The farther left formerly represented by Podemos is running in a new alliance called Sumar, while the far-right Vox may finish in third place. A key question for the election is whether PP+Vox will be a majority and if so, will they govern together (either in coalition or in confidence-and-supply)?

One thing I always watch in Spanish elections is the inevitable differential treatment of parties by the electoral system, which can be consequential for seat shares and thus who can govern. There are four major national parties/alliances in this election, where “major” loosely means expected to get over 10% of the vote. And there are many regional parties. District magnitude is highly variable, with many districts having five or fewer seats (mean is 6.7), and D’Hondt divisors are used. It might be thought of as the ultimate example of districted PR, as few other PR systems have so many districts or such a range of magnitude (1 to 37). So the regional distribution of votes matters greatly to the overall outcome.

A good way to follow results as votes are counted is at results website of the Ministerio del Interior; they also have a good smartphone app (just search on “elecciones España 2023” or scan the code at the bottom of the just-linked site).

On ‘The Tally Room” talking about the Seat Product Model

For the second time in 2023, I have made a foray into podcast land. I was interviewed by Ben Raue for his podcast, The Tally Room: “From seats in to votes,” in which we discuss what the Effective Number of Parties is and how the Seat Product of an electoral system shapes party systems. We discuss Australia extensively, but also Canada and various other systems. Even St. Kitts & Nevis gets a reference.