Two prominent cases of MMP that suffered from “decoy parties”: Albania 2005 and Lesotho 2007

Posted on 15/03/2026 by msshugart

The following is a brief I wrote for a group interested in the possibility of an MMP system for the US. (More on that some other time, perhaps.) In our discussions, the topic came up of how to deal with potential “decoy” parties, whereby a party can “game” the system by encouraging supporters to vote for one party in their constituency vote and a formally distinct party on the list, bypassing MMP’s compensation mechanism. There have been two prominent cases where this has happened, Albania 2005 and Lesotho 2007. (The latter case occasioned a long and interesting comment thread here about how to prevent this.) The following focuses in part on the question of whether one could prevent the decoy problem by requiring a party that presents a list to have constituency candidates, or vice versa. I am skeptical of that, and more so after diving into the district-level data for those two elections.

The previous planting looked at “strategic overhang,” where a small party seeks to win seats beyond its party-vote entitlement, thereby increasing its leverage over a potential ally in government formation–which might occur with or without the cooperation of the larger (potential) partner. Here we will be concerned with a major party strategically setting up a formally separate party, where one focuses on winning districts while its “decoy” focuses on winning list seats.

Albania 2005

This case featured a lot of parties running in an alliance, so I will focus on the main ones.

The Democratic Party of Albania (DPA) won 56 of the 100 constituency seats on 44.1% of the aggregate vote for constituency candidates. The next largest party was the Socialist Party of Albania (SPA), which won 42 constituencies on 49.4% of the vote.

The DPA had a bunch of decoys! The main one was the Republican Party of Albania, which won 11 of the 40 seats on about 20% of the list vote. Importantly for our purposes, it presented no constituency candidates. Nor did any of the other decoys. There seem to have been 6 such parties, but only three others aside from the Republicans won any seats. Total, these allied list-only parties won 18 seats.

Also importantly for our purposes, the DPA actually also presented its own list. But it got only 7.7% of the list votes and no seats. I can only guess that the law required a party with constituency candidates to present a list, so they dutifully did. (Maybe someone can confirm.) But they directed the voters over to the decoy parties’ lists. The allied party lists amassed 41.1% of the votes, so just a little behind the DPA itself. The DPA won no seats for its own list, even though it had 7.7% of the votes, while other parties with smaller list-vote percentages won seats. Of course, that is as expected, because it was MMP! It was already over-represented via the constituency side, so it earned no compensation seats. But its decoys allowed it to get those other 18 list seats, when in the absence of the gaming it probably would have won only one or a few list seats.

In the end, this resulted in the DPA and its allies combining for 74 of the 140 seats, despite only 44% and 41% of the vote (constituency and list, respectively) and the supposed full compensation of MMP.

It seems that the SPA also played the game, in that their list had only 8.9% of the vote. There were several other parties with “Social” or “Socialist” on their names that got significant list vote totals (and seats) but got few votes in the constituencies. But I am not sure which were decoys and which were “legitimate” parties. See appendix for the full set of parties that ran constituency candidates or lists. (The “appendix” will be in a comment, with apologies for the formatting challenges.)

Don’t miss the very helpful additional information provided by Etjon Basha in a comment!

Lesotho 2007

Main parties were Lesotho Congress for Democracy (LCD) and the All Basotho Convention (ABC).

Neither of these parties ran a party list in this election.

The LCD’s decoy was the National Independent Party. The ABC’s decoy was the Lesotho Workers’ Party. The National Independents had at least one constituency candidate (earning 0.64% of the total constituency votes cast in the country). The Lesotho Workers’ Party had no constituency candidates.

In the result (see at Wikipedia), The LCD won 61 of the 80 constituencies (on 52.5% of the aggregate constituency vote) and the ABC won 17 seats (29.3% of votes). There was one other party winning a constituency seat.

On the party-list side, the National Independent Party won 21 of the 40 seats on 51.8% of the list vote. Note it was only a little below what its “parent” party, the LDC got. The Lesotho Workers got 10 seats on 24.3% of the vote. Note they did about five percentage points worse than their parent.

The result shows the “LCD Alliance” having won 82 of the 120 seats despite around 52% of the vote. So they had a two-thirds majority, when proper functioning of MMP would have implied about 64 seats. In other words, they should have won around 3 list seats to “top up” their 61 constituencies, instead of the 21 list seats their decoy won. The “ABC Alliance” ended up with 27 seats, which shortchanged them. Thus they did not draw any benefit from the decoy strategy, but only because they lost the voting contest so badly, and the LDC’s decoy had taken more than half of the list seats.

Solutions

The two countries adopted different responses to this problem of decoy lists. In Lesotho, they abolished the two-vote feature, keeping MMP to this day, but with each party’s entitlement to compensatory seats determined through an aggregation of all of its candidate votes. In Albania, they abolished MMP entirely and shifted to a purely list-PR system (initially closed lists, later open).

–compiled by Matthew Shugart, February 2026, from The Elections Archive (THEA), Wikipedia, and various other sources.

Strategic overhangs and ways to prevent them in MMP systems

Posted on 15/03/2026 by msshugart

In mixed-member proportional (MMP) electoral systems, an “overhang” seat can result if a party wins more seats via the nominal tier (of, typically, single-seat districts) than it would be entitled to if a regular proportional-representation (PR) system were used (i.e, one lacking a nominal tier, whether proportionality is jurisdiction wide or calculated region-by-region). Sometimes these excess seats come about simply because one of the parties happens to be strong enough to win a lot of district races–usually decided by plurality–against a fragmented set of opponents. Such a case occurred in the recent election in Baden-Württemberg. Sometimes one or more overhangs occurs because some small party wins one or more district contests but gets such a small list-vote share that it would not have won so many seats (or any at all) under a standard PR system. Such cases have happened several times in New Zealand and a few times at the federal level in Germany,

The case of a party dominating the nominal tier is not a strategic choice by the party, at least not in the sense that I would understand party strategy. It is not really the party’s fault if it happens to have broad enough support to win a lot of districts. However, the case of a smaller party winning overhangs might be strategic. That is, it might be seeking district wins as part of a plan to boost the seat total of a group of two or more parties beyond what the bloc’s actual support level is in terms of the list votes. A party can do this by entering candidates in one or more key seats where it might win, while encouraging its voters to give their list vote to some larger allied party. Many critics of MMP would say the this sort of strategy is a “rort”–a way of manipulating the system for advantage by undermining the system’s design goal of proportionality. (I am not necessarily making that argument myself, and there may be cases that are manipulation and others that are not, by whatever standard we might choose to apply. For one example much discussed at the time, see the “Tea tape scandal.”)

There is, of course, another type of strategy that I think almost anyone would agree is manipulation–the case where a major party sets up a “decoy list,” and encourages voters to vote for its district candidate and its decoy list. This strategy, if successful, actually converts the MMP system into a majoritarian system: the big party wins its typical large haul of the nominal seats and the decoy wins seats off the list as if it were an independent party. Because the main party and its decoy are technically separate parties, the decoy wins compensatory seats that are not limited by the “entitlement” of the main party, but rather as if it was a party that failed to win district seats and therefore is entitled to a large chunk of of the compensation tier. This has happened in one election in Lesotho (which occasioned a long and interesting comment thread here about how to prevent this), and one in Albania. It also has happened in South Korea recently, although that system is not strictly MMP.¹ (I am going to put up a separate new post with some detail on these incidents in Albania and Lesotho.)

Whenever MMP is proposed, among the questions that come up are how you cope with overhangs (the preceding post deals with this question in some detail), and how can you prevent strategic manipulation of the overhang provision?

A post by Craig Rikihana from last December caught my attention on these points for sketching out a case of a potential “overhang gambit” in the upcoming 2026 election in New Zealand. (If you are not familiar with how MMP works or how it came about in New Zealand, he offers useful overviews of these points as well.) The “gambit” concerns the Maori Party and the special constituencies within the system’s nominal tier in which voters who register on the separate Maori voter roll cast their district votes. The party leader is hoping to win up to six of the Maori seats,² “while deliberately keeping the nationwide party vote for Te Pāti Māori low.” Rikihana goes on, “This alters the mathematics of government. A major party needing TPM’s seats to reach a majority in a 123-seat Parliament (where 62 are needed) must negotiate from a position of weakness. The overhang makes TPM’s support more costly and valuable.”

Something like this has happened before, although perhaps Rikihana is claiming it was somehow less deliberate or less a “gambit.” For instance, in 2008, the Maori Party won five electorate seats but its 2.4% of the list vote would not have entitled it to any seats had it not won at least one nominal-tier seat. The threshold in New Zealand is 5% of the party vote or at least one electorate plurality. By winning one electorate, a party is then eligible for full compensation if it has a party vote sufficient for additional seats; 2.4% would have entitled the party to three seats (again, conditional on having won an electorate). However, by doing so well at the electorate level, it won five. Under New Zealand’s procedure for dealing with overhangs, this meant adding seats to parliament so that other parties do not have their total number of seats reduced. Parliament in 2008 thus increased from its baseline 120 seats to 122. National, the major center-right party, won a total of 58 seats and its main ideological ally, ACT, won 5. These 63 seats actually would have been sufficient for a majority regardless of the Maori Party overhangs. The latter party’s five seats, however, made it an attractive additional partner for National (reducing the leverage of ACT) and it was brought in as a party having “ministers outside cabinet” as is often practiced in New Zealand since MMP was adopted.

If we were to condemn the ambitions of the Maori Party to win multiple overhangs and the potential thereby to make itself pivotal through such a strategy–and again, I am not engaging in any such condemnation–then we might ask what could be done to prevent it. Rikihana claims this gambit is “unique” to New Zealand. He says that other MMP systems “add “balance seats” to fully compensate for overhangs, neutralising any leverage.” It is true, of course, that Germany’s recent MMP systems have allowed the Bundestag to increase greatly in size if needed to restore full proportionality. This was also the case in the recent Baden-Württemberg election, as outlined in the preceding planting. (Remember that this is no longer done at the federal level in Germany, and in fact due to other changes, that system no longer should be considered MMP.)

But is he right about Wales and Scotland?³ I am not aware of any such provisions. In Scotland, every election since the current Scottish Parliament and its electoral system were established has had 129 seats, 73 single-seat districts, and 56 compensatory list seats (determined in regional compensation districts, not Scotland-wide). Overhangs are actually quite common in Scotland, as I define the term. Again, an overhang is any seat a party wins in the nominal tier that puts it over its proportional entitlement. (It is not the seat added to an expanded parliament to counter any such overhang, if the rules call for such a procedure, although the term is sometimes used that way.) Scotland actually has often seen overhangs, inasmuch as the Scottish National Party has been so strong at the district level that it often wins considerably more seats than it would win if the regions were multi-seat PR districts with no single-seat constituencies within them. For instance, in 2011, the SNP won over 70% of the constituency seats on 45.4% of the nominal vote. The party also won 16 lists seats (because there were compensation areas where it had been underrepresented within that region), so it had 69 of the 129 seats (53.5%) despite having won only 44.0% of the party list vote. No “balance” seats were added to prevent such a substantial overrepresentation.

Similarly, every election in Wales under its former MMP system had exactly 60 seats elected. No seats have ever been added to deal with overhangs. As recently as 2021, the leading party (Labour) won 27 of the 40 constituencies despite only 36.17% of the list votes, an obvious overrepresentation that was compensated only to the extent possible under the fixed number of seats available from the list tier. (Similar to Scotland, in Wales the compensation is carried out in regions; Labour also won 3 list seats in 2021.) It should be noted that Wales will no longer use MMP in the upcoming election.

This factual error about “balance” seats in Scotland and Wales notwithstanding, I do not know if there have been cases of “overhang gambits” in elections in these jurisdictions. We had a discussion about some potential “gaming” prior to the 2021 election in Scotland. I was doubtful that the case counted as manipulation, but it did raise some eyebrows. The party that led to the concern, Alba, ended up winning no seats. So if it was a gambit, it was a failed one.

If readers know of other cases in either Wales or Scotland (or London, which also has MMP) in which there was an “overhang gambit,” successful or otherwise, I’d be grateful for the primer. Also, do these jurisdictions have any specific provisions in law or administrative practice that would prevent strategic overhangs, decoy lists, or other obvious or alleged manipulations? It is not clear precisely how one would prevent this sort of alleged manipulation of MMP that comes from small parties seeking seats beyond their proportional entitlement. Such strategies are perhaps inherent in the system. As I have alluded to with respect to the Scottish case, it is not even clear that this sort of strategic overhang is an actual problem needing a fix. In that sense, it is different from the issue of decoy lists, which do seem like a straightforward manipulation, but also one that may have fixes that can be implemented within the context of MMP rules.

In his blog post, Rikihana claims that in Lesotho “overhangs are common and uncompensated” but are “incidental,” rather than a “crafted strategy.” This is, of course, incorrect with regard to one specific election. The case in Lesotho in 2007 was a classic decoy-list strategy. Its occurrence led to the two-vote MMP system being changed before the next election to a single-vote version: now there is no separate list vote, but instead the votes for candidates are accumulated across districts to arrive at list totals for purposes of determining compensation seats. This is a different response from the one taken in Albania, where MMP was abolished entirely in favor of a simple districted list-PR system, since 2009.

Because this post has already become rather long, I will make a separate one with some details about the decoy list strategies in Albania and Lesotho.

_____

There is also the case of a former system in Italy, used in 1994–2001, but that also was not MMP, and the setting up of allied parties there was occasioned by a specific feature of that system. ↩︎
The total number of such seats depends on how many voters opt to register on the Maori roll. In 2023 there were seven Maori seats. ↩︎
Terminology note: In Wales and Scotland, what I would call a variety of MMP is referred to as an “additional member system.” I dislike this term and do not use it. The list seats are not “additional”; they are compensatory. ↩︎

Tie for coalition parties in Baden-Württemberg 2026, and the varieties of MMP

Posted on 12/03/2026 by msshugart

The German state of Baden-Württemberg held its state legislative election this past Sunday, 8 March 2026. The incumbent government is headed by Winfried Kretschman of the Green Party in coalition with the Christian Democratic Union (CDU). The latter is the party of the federal Chancellor Friedrich Merz. The government following this election is sure to be a continuation of this two-party coalition, now with new Greens leader Cem Özdemir at its head.

In addition to reviewing the election outcome under the mixed-member proportional (MMP) rules in place–which were reformed prior to this election–this planting will also consider how some alternative rules, including other variants that remain MMP but are not as fully proportional, would have worked.

In the voting result, the two parties were almost tied, with the Greens on 30.2% and the CDU on 29.7%. In third place is the far-right Alternative for Germany (AfD) on 18.8%, followed by the Social Democrats (SPD) on a mere 5.6%. The legal threshold for winning any list seats is 5% of the party vote, and no other parties cleared it. This year’s result is actually a substantial gain for the CDU compared to the state’s election of 2021, when it had 24.1%, and a small decline for the Greens, who had 32.6% in the preceding election. These vote figures are not fully comparable across years, for reasons having to do with the electoral reform, discussed later.

A table at Wikipedia has the full 2026 results, broken down by list and nominal (constituency) vote, given the mixed-member proportional (MMP) system in use. Interestingly, the CDU did quite a lot better in the nominal vote, with 34.3% to the Greens’ 25.5%. Unlike the reformed national system, the B-W system still has the possibility of increasing the total size of the assembly in the event that there are “overhangs” from a party having earned more constituency seats than what its proportional entitlement should be. In addition, like the former federal rules, the B-W rules still allow for the possibility of further “leveling” seats to restore full proportionality (for above-threshold parties).

In nominal seats, the CDU won 56 of the 70 total. The assembly has a minimum size of 120 seats, implying 50 list seats. According to the brief description at Wikipedia, the law requires proportionality statewide. The CDU’s 56 constituencies won would already be almost 47% of the minimum 120 seats, which would be a substantial overrepresentation given their 29.7% of the list vote (which, as noted, was a second-place finish). Thus the CDU is entitled to no list seats, given that MMP mandates these be allocated as compensatory seats (not in “parallel” as in the other main type of mixed-member system, MMM–see Japan’s recent election, for example). [Deleted from this paragraph was an earlier point about proportionality also being required across administrative district. See comment by Thomas for correction.]

Other parties, however, need compensations seats, and the mandated full proportionality can be achieved only by a substantial increase in the size of the assembly, which will have 157 seats. In the final result, the Greens win 43 list seats in addition to the mere 13 districts they won. This total of 56 puts them on par with the CDU, whom they essentially tied in party-list votes. It also leaves each of the two large parties on 35.7% of the total expanded assembly. Given that 15.75% of votes were cast for parties below the legal threshold, the Greens and CDU each have approximately 35.6% of the above-threshold vote. Thus indeed full proportionality has been restored through the allocation of the expanded pool of compensatory list seats.

As for the other parties, only one nominal-tier seat was won by a party other than the top two. This was an AfD seat, and that party’s 18.8% of the total vote translates to 22.3% of the above-threshold vote. It has been allocated 34 list seats to go along with its one constituency seat, and that total of 35 is, yes, 22.3% of the total 157. The only other seats allocated are 10 list seats to the SPD.

Now I want to turn to potential alternative ways in which seats could have been allocated and still be consistent with the principles that define MMP, as well as one that would not be.

The system would still be MMP, albeit less fully proportional, even if there were no expansion of the size of the assembly to deal with the overhang. If the number were fixed at 120, with 70 single-seat constituencies and 50 compensatory list seats, the CDU’s 56 constituency victories would have been 46.7% of incoming assembly. The three other parties that cleared the party-vote threshold would have split the 50 list seats. There are various ways this might be calculated, but a straightforward way if we want a simple estimate* is to recalculate list vote shares for only those parties that are entitled to list seats. The parties that are not so entitled obviously include those that are below the list-vote threshold. Also included among such parties are any that already have obtained at least their proportional entitlement in the nominal tier. In our case, that means we exclude the CDU votes in addition to all those below the threshold. When we do this, the recalculated vote shares of parties eligible for list seats are: Greens 0.554, AfD 0.345, and SPD 0.102. These shares are applied not to the total 120 seats, but to the number of seats remaining to be allocated, which in this case is 64: our fixed number of 120, minus the 56 already credited to the CDU from the nominal tier. When we apply these adjusted vote shares for parties eligible for list seats, we get approximately¹ 35 seats for the Greens, 22 for the AfD, and 7 for the SPD. Some of these parties already have nominal seats, so their list seats will be these adjusted entitlements, minus nominal seats: Greens 22 (because they won 13 districts), AfD 21 (they have one nominal seat), and finally the SPD which gets all of its seats from the list (7). The resulting percentages of the full 120 seats would be: Greens 29.2%, CDU 46.7%, AfD 18.3%, SPD 5.8%. [*See footnote #1 (second paragraph) for a more detailed way that is probably more accurate.]

Recall that the Greens obtained 30.2% of the list vote, which was a plurality. They are somewhat underrepresented in this scenario, with an advantage ratio (%seats/%votes) of only 0.966. The CDU’s 56 seats give them an advantage ratio of 1.57, and those of the other two parties are 0.975 (AfD) and 1.05 (SPD). These differential advantages ratios (1.00 would be pure proportionality) are due to the overhang, as well as the lack of any seats to be added in response to those overhangs. This is assuredly MMP, and the result is far more proportional than if there were only single-seat districts (the CDU, after all, won 80% of those!) or if it were mixed-member majoritarian (under which the CDU still would have won list seats–probably 15 of the 50) despite having won so many districts.

An intermediate approach between the model just sketched and the actual procedure would be to add seats to the list tier for each overhang, but no further “leveling” seats beyond those. This would be similar to how the New Zealand MMP system works (see in 2023 or 2008). In this scenario, because the CDU has 14 overhangs (as we shall see), the assembly would be increased from 120 to 134 seats. The resulting 64 list seats would be used to compensate the other parties. Under this procedure, each party that does not have an overhang still gets the same number of seats it would have had in the absence of the overhang. Given the parties’ shares of the above-threshold vote, the seats should be, under the no-overhang scenario: Greens 43, (CDU 42), AfD 27, SPD 8. The CDU’s share shows us where the 14 overhang seats come from, given it actually won 56 single-seat districts. Because of the nominal seats won, these imply list seats for these parties of Greens 30, (CDU 0), AfD 26, SPD 8. Note that these sum to 64 list seats–the baseline of 50, plus the 14 added in response to the overhangs. In terms of seat percentages (and advantage ratios), we now have Greens 32.1% (1.06), CDU 41.8% (1.41), AfD 20.1% (1.07), and SPD 6% (1.08). All parties are now slightly overrepresented (advantage ratio over 1.00), but the CDU remains substantially more overrepresented than the others. The MMP system still effectively rewards it for having many candidates who could win districts, yet it does not cut any other above-threshold party’s number of seats from what it would have been in the absence of overhang. The shares are still reduced relative to the no-overhang situation, because there are no further “leveling” seats added to restore full proportionality, as there are in the actual system in use in B-W.

And then there is yet another possible solution to the overhangs. The CDU could be required to give them up, and the 14 weakest of their 56 districts would be left without local representation,² leaving full proportionality to be achievable for all above-threshold parties out of the original 120 seats. Note that this is utterly inconsistent with the principle behind MMP, which is that candidates hold seats if they have been able to win the local plurality, and parties win proportional representation up to the limits of the total seats available (expanded, as necessary and permitted under the rules). This is the reason why I no longer consider the federal system of Germany to be a mixed-member system. It violates the key principle that the nominal tier results are decisive. It is instead a form of PR with constituencies serving as a local-nominating and overall intraparty ranking mechanism. B-W, however, retains a very pure form of MMP, with maximum proportionality (for above-threshold parties) as its design goal.

As alluded to earlier, this system is itself a recent reform, passed after the 2021 election. B-W used to have no separate party list. Instead, the compensation seats were awarded to the best-performing constituency candidates who did not win local pluralities. Again, this is fully consistent with MMP, and we could say simply that the compensatory “list” is composed of constituency-nominated candidates who are non-winners by that tier. (The 2021 system also had added compensatory seats in case of overhang and additional “leveling” seats to restore proportionality, like the 2026 system. That is, how the list is constructed and whether there is a separate vote for it are separate dimensions of MMP from how the system deals with overhangs.)

Because there was only a single vote in 2021, arguably a direct comparison of vote changes between these elections is not straightforward. In 2021, a voter could vote for the “list” of a preferred party only by voting for the party’s constituency candidate. If we compare nominal votes for constituency candidates in 2026 to the 2021 fused vote, the CDU gained over ten percentage points (34.3% from 24.1), while the Greens fell by 7.1 (32.6% to 25.5%). If we compare across list votes it is +5.6 for CDU and –2.4 for Greens. Which is the better comparison? Arguably we can’t know. Voters under single fused-vote MMP must make a choice between endorsing their preferred local representative or their preferred party (if these are different for a given voter), and that means some may vote for a a small party with no chance of winning their district in order to boost its overall seat total via compensation. Others may vote for a candidate even if they do not really like the party, simply because they want that local representative. This is why I think one-vote MMP undermines the “best of both worlds” potential of MMP! However we look at the cross-election change, the CDU gained and the Greens lost, yet both will remain in coalition and the Greens presumably will retain the state’s Minister-President position on account of their narrow plurality of the separate list vote.

The voters indeed did take advantage of the ticket-splitting option. The effective number of vote-earning parties was 4.50 in 2026 in list votes, which is higher (albeit not by a lot) than the effective number on nominal votes, which came in at 4.34. The effective number on seats is a good deal lower, due to the large below-threshold vote, at 3.24.³ In 2021, the effective number of parties via votes was 5.01 and seats 3.89. The 2026 result is less fragmented (lower effective numbers) because the votes and seats are more dominated by the top two parties, even if those two combined for not quite 60% of the total votes and barely over 70% of the seats.

Regarding the legal threshold, two parties obtained over 4% of the list votes but below the mandated 5%. The Left Party won 4.41% and the Free Democratic Party (FDP) won 4.38%. No other party won over 2%. As I have noted before, it is quite possible that the 5% threshold is now too high for the party system that Germany and its states actually have, and that it should be lowered to 4% or even 3% or lower to avoid such a high total of wasted votes.⁴ The wasted votes were close to a sixth of the total, and would have been cut by more than half had the threshold been 4%. Having the Left and FDP in this assembly almost certainly would not have changed what governing coalition could form based on this result. But in other cases, it could have such an effect.

I always enjoy looking at the list of also-rans, the parties that fall far below a legal threshold. In this election, my favorite might be the “Party for Rejuvenation Research,” which earned 3,590 list votes (0.07%).

The Baden-Württemberg election of 2026 results in a tie in seats for the top two parties, Green and CDU, because (1) those parties had almost tied list-vote shares, and (2) the rules of MMP in use provide for additional seats to correct for overhang and further additional “leveling” seats to restore full proportionality for parties that clear the list-vote threshold. There are, however, other potential means of handling such a large overhang as the 14 CDU seats that were above its proportional entitlement, and two of those that we have reviewed remain within the parameters that define MMP. A third one we reviewed–which would be similar to that used since 2025 in federal elections in Germany–would be outside the family, as strictly defined. Baden-Württemberg is one of several German states holding elections in this Superwahljahr. Most German states use MMP, with some variation in the details of system design that we may explore further as these upcoming elections occur.

_________

For ease of calculations, I am applying the adjusted vote shares directly to the number of available seats and rounding. The method used in the actual allocations is Sainte-Laguë divisors, generally considered one of the most “proportional” allocation formulas. Thus my shortcut method should be okay for illustrative purposes. However, running the full algorithm could produce marginally different seats totals by party than what I report here.
Later, I did a full Sainte-Laguë allocation that produced slightly different numbers: list seats by party as Greens 24, CDU 0, AfD 20, SPD 6. The updated seat percentages (and advantage ratios) by party are then 30.8% (1.02), 46.7% (1.57), 17.5% (0.93), 5.0% (0.90). For this revised calculation I actually ran the full divisor set on the vote totals of all above-threshold parties, with the initial divisor being the number of nominal seats won, plus 1. This, of course, cuts the CDU vote total down far enough that it can’t win one of 50 list seats, and ensures that each of the others obtains its compensation share. This should be more accurate, but note that the initial method I used was a lot faster and differed by two seats out of the 50. Not too bad. If you want to see what the table of quotients looks like, click here [Dropbox link to a PDF]. The bolded quotients in the table are the ones that were sufficient to earn a seat for the indicated party. ↩︎
Or those districts could be represented by the candidate who came second in the nominal contest (provided that party’s list was eligible for seats), without changing the overall number of seats this or any other party would win. However, the current federal system simply leaves the district without an officially designated constituency member. ↩︎
By my calculation (sketched earlier), had there been a rule allowing for no additional compensation seats in the event of overhangs, the effective number of seat-winning parties would have been only 2.92 (2.89 by the revised full allocation referenced in footnote 1). ↩︎
See the final paragraph of the linked post. ↩︎

The Esparto explosion site, almost eight months later

Posted on 23/02/2026 by msshugart

On the first day of July, 2025, our local area was rocked by a series of loud explosions and a fire at what turned out to be a fireworks facility operating outside the law. Yesterday I took several photos of the site. Not much has been done with it. Even debris in an adjacent field remains. This is, of course, a crime scene and the investigation is ongoing. The photos show the intensity of the fire and force of the explosions.

KCRA put together a video with a timeline of the events on the scary evening. In an earlier post I related my own experience.

Click on any photo to see a larger version. (No captions; they kind of speak for themselves.)

Frost-protection measures, Feb. 2026

Posted on 22/02/2026 by msshugart

As stone-fruit bloom was just getting underway about three weeks ago, I mentioned my concern that with such an early bloom, we could still have a bout of freezing weather and/or frost. And sure enough we did. The forecast was not for such a degree of cold that I was overly worried about the stone fruit blooms or early fruit set (and there was certainly no low snow, as there was almost exactly three years ago). Rather, I was concerned with the younger citrus and avocado trees that also had taken advantage of the warm spell to begin their new growth, including flower buds. So I took some protective measures. Not as drastic as what I have done other years with more extreme conditions forecast, but I put up some strings of warming lights and bedsheets to try and keep the temperatures on the tender new growth a little warmer than the ambient air.

The first image shows a covered Tahitian pummelo in the foreground and two avocado trees (Stewart and Zutano) in the center. Because it was not forecast to be much below 30F, I did not bother to bring the sheets over the avos all the way down to the ground for a full enclosure. My objective was mostly to keep frost off the buds as well as to hold in a little of the heat generated by the lights strung around the trees (note the cords as well as the glow of some of the bulbs through the sheets). As we will see below, this was successful. The second photo shows a shroud over a young Cara Cara navel orange. In the foreground at the left is a Washington Navel orange, uncovered, with some of its ripe fruit visible. This is a relatively large tree, which would be challenging to cover, and as a more mature tree it would be unlikely to threatened unless it would get much colder. Farther up towards the house are some other young citrus trees over which I also put lights and a sheet.

How much difference to the temperatures did these actions make? I have sensors for that, and hence data! Below are the reports over the two nights of cold temperatures from two sensors. “Sensor 2” is in the open near one of my citrus trees. “Sensor 3” is hanging inside one of the avocado trees, and thus was under the sheet where there were lights on during the night. No light was anywhere near direct contact with the sensor, so it should have given a good reading of the nighttime air temperature beneath the canopy of the sheet (which in turn was supported by stakes).

I put the box with time and temperature reading for each sensor at about the point that the lowest value was obtained on the morning Feb. 20. The air temperature in the general area was 30F at 6:30 a.m. Around the same time, it was just over 35F inside the canopy where there were lights to warm the air. Thus my technique appears to have provided the trees with about five degrees extra warmth. That is even better than I would have expected. I also noticed no frost on top of the sheet over the trees that had strings of lights.

On another tree that I covered with a sheet but did not have any lights on, I noticed a bit of frost on top of the sheet. See below, viewed from inside the house at about 8:15 that morning. This is my Wekiwa tangelolo, one of my prized citrus trees. I know from experience with a hard freeze in December, 2013, that this tree is tough. Even though it was newly planted then and not well protected, it survived several hours at 20F. It had severe damage and defoliation, but recovered well. I know I do not have to take aggressive protective action for it unless the temperature will be very cold. Still, I am glad the frost settled on a sheet (which in turn is on a wooden frame like what you can see over the jaboticaba behind the Wekiwa) instead of on the tender green shoots. (I got lucky in 2013; the forecast was not for it to be that cold but even so I was not prepared like I should have been.)

On the morning of 22 Feb. I removed the sheets, because rain is expected in a couple days and I do not want the extra weight of waterlogged sheets on the trees. Plus, I wanted the trees to enjoy some sunshine today. I will leave the lights strung over the trees for now (unplugged, for safety), just in case I need them again in the coming weeks. The lights were not on over the night of 21–22 February, given it was not forecast to be much below 35. So this allows a chance to see how much difference just a sheet but no added warmth from lights can make. The graph below shows the two sensors again, with temperatures at sensor 2 on top and sensor 3 on the bottom (ignore the humidity reading for sensor 2 in between). The result is that the sheet might actually have trapped a bit of cold air, as it was actually colder inside the canopy where sensor 3 was hanging than it was at sensor 2 in the open, although it is a minor difference. It was about 0.4 degrees warmer in the open, presumably because the sheet inhibited mixing of the air from whatever light breezes we may have had overnight.

Also of interest to me is what difference the sheets might make during the day. The answer is quite a lot! The graphs above show it was around 8.4 degrees warmer in the daytime on the 21st under the sheets! I would not have expected that much, although I certainly know not to leave these on if a really warm day is expected between two freezing nights (not a highly likely event, although 30-degree day–night swings do happen here even without sheets or other material intervening). We can also see that when I removed the sheets, initially the temperature at sensor 3 dropped rather substantially before daytime heating brought it back up. See the dip in the bottom graph, towards the right, between 9:45 and 10:30 at a time when the sensor 2 temperature reading was still rising steeply. (The later dip at sensor 2 is likely due to either cloud cover or advancing shade from the house, whereas the later rise at sensor 3 is because this sensor gets some direct sunlight on it by midday–when there is no sheet over it, that is.)

All in all, a worthy little experiment in frost protection. I will leave you with a picture of some frost on my curry leaf tree. This tree is tough and always recovers, but will suffer some minor damage from even a light frost.

Costa Rica 2026: A sudden drop in the fragmentation trend

Posted on 20/02/2026 by msshugart

In Costa Rica’s election on 1 February, Laura Fernández Delgado, candidate of the Sovereign People’s Party, won in the first round with 48.6% of the vote. This party was founded only in 2022 and is considered by some to be a right-wing populist party. The party also won a majority of seats in the concurrently elected Legislative Assembly, with 31 of the 57 seats.

Costa Rica elects its president via a qualified plurality formula, specifically 40% of the vote is required and if no candidate reaches this threshold, there is a runoff between the top two. Fernández not only reached it easily, she won by a wide margin, with the runner-up (Álvaro Ramos Chaves from the venerable National Liberation Party, or PLN) way back on 33.6%. The PLN has won 17 seats, the Broad Front 7, and two other parties a seat apiece. These one-seat parties are Coalición Agenda Ciudadana and Unidad Social Cristiana (PUSC). The latter is noteworthy in that it shows, as of this election, how the old right has been supplanted by the Sovereign People’s Party. The PUSC was one of the two top parties, along with the PLN, from the mid-1980s until the 2000s, and has provided three of the country’s presidents, although the last of these was elected in 2002.¹

For most of this planting I am going to focus on the extent to which this result represents an abrupt halt to the country’s recent rather extreme party-system fragmentation. The graph above shows the values of the effective number of parties (assembly votes and seats) and presidential candidates for all elections since 1953. Costa Rica had long been known for having two dominant parties and a few smaller ones. I wrote about the fragmenting trend following the 2022 election. In that year’s assembly election, the effective number of vote-earning parties reached almost 8.0,² which is right up there with the highest values observed at any time, anywhere in the world (more formally, it is in the 99th percentile). On seats, the effective number in 2022 was 5.02, which is also quite high (though merely 75th percentile), but especially high by Costa Rican standards. And then in this year’s election it suddenly has come all the way down to just short of 2.5. The effective number of seat-winning parties thus was cut in half in a single election. As we can see, the effective number of vote-earning parties was cut even more dramatically. In fact, it can be considered a world record for one-election defragmentation in any established democracy.

I was curious where Costa Rica would rank among established democracies in terms of its previous major fragmenting trend as well as its now record single-election drop in fragmentation. This is the sort of question for which I collect all that data, so let’s see!

There are obviously various ways one could operationalize the concept of major fragmenting trend. I was specifically interested in what precedent there might be for a democracy to go from a relatively consolidated party system (let’s say an effective number well below 4) to a high number (like 8 or higher). I chose to operationalize this by establishing mean values for all democracies through the year 2000, and then look at differences between values obtained at subsequent elections and the established pre-2000 trend in the same country.

For countries that had at least ten elections through 2000, Costa Rica’s 2022 result is the third greatest increase for any country that was previously below 4 in its effective number of vote-earning parties (N_V). The country’s mean N_V was 2.83 for the pre-2000 period and hit 7.99 in 2022. The two bigger increases from such a low base are Colombia, with its mean of 2.68 prior to 2000 and a high of 9.7 (!) also in 2022, and Greece which went from 2.80 to 8.95.³ Colombia’s difference over its prior longterm mean was +7.01 and Greece’s +6.15, whereas Costa Rica’s is +5.16.

There actually are higher differences in the dataset, but from already-high bases or with the base being over fewer elections. Brazil went from a mean of 8.82 to a high of 17.97 in 2018, for a change of +9.15. However, that relatively lower mean was over just three elections, and in any case was already extreme fragmentation. Just not as extreme as it would become. Also worthy of note is Belgium, which went from a mean of 6.09 (averaged over 18 elections) to a high of 10.96 (2019⁴). That is an increase of “only” 4.87 but still notable. The Netherlands also gets an honorable mention here, from a base of 4.98 to a high of 9.26 in 2021. As regular readers or election-watchers know, the Netherlands 2025 election saw extraordinary fragmentation, although the effective numbers were not quite as high as in 2021.⁵

So, with that information in hand–Costa Rica is the third most extreme case of a fragmenting trend by the definition adopted–I wanted to know how this 2026 result compares among the all-time single-election decreases in fragmentation. Voting totals for the assembly election are still not official, but the preliminary results will be good enough. The Sovereign People’s Party, the party of the incoming president, won 46.3% of the vote. Note that is roughly 95% of what the presidential candidate obtained, which is fairly typical for concurrent elections. While farther short of a majority than the presidential vote total, it nonetheless translated into a manufactured majority of legislative seats. While Costa Rica uses a proportional-representation electoral system, it is one with a relatively low seat product (464), which is what enables a seat majority on only around 46% of the vote,⁶ with an assist from a rather substantial falloff from largest to second largest party in votes.

The runner up in the assembly vote was the PLN, with 23.9%, which is only around 71% of what the PLN’s presidential candidate obtained. Thus the assembly voting remains somewhat more fragmented than the presidential, which of course is also not unusual for concurrent elections, especially where some form of proportional representation is used for the assembly.⁷ The top two combined for 82% of the presidential vote, but only 70% of the assembly vote.

As noted above, the effective number of seat-winning parties came down to around 2.5 in this election. The effective number of vote-earning parties (N_V) remains considerably higher, at 3.42. However, this is less than half what it was in 2022, when N_V=7.99. So this change of –4.57 is what we need to answer our question about all-time single-election declines in fragmentation in an established democracy. This single-election decline appears to be a new record! For a country that had at least ten elections through 2000 (to keep the parameters the same as the earlier question about increases off a long-term trend), the previous record would seem to be Israel in 2003, when N_V fell by 2.92 from the preceding election (1999: N_V=8.69; 2003: N_V=6.17). In fact, it demolishes this record! Moreover, Costa Rica’s change happened under steady-state institutions, whereas Israel’s previous record for a long-term democracy was set following a significant change. In 1999 in Israel, the prime minister was directly elected, whereas in 2003 there was a reversion to pure parliamentarism.⁸

Considering all elections in my dataset, there are two bigger one-election drops in N_V but they occurred in countries that were not, at the time, long-term democracies. Poland had a one-election drop of –5.21 in 1997 (only its third democratic election), and Japan had –6.04 in 1952 (the 4th election after WWII).

So Costa Rica, with the 2026 election, has swung from having the third-greatest votes fragmentation trend among countries that had established democracies as of 2000 to the record single-election drop in fragmentation ever recorded in a long-term democracy. Although less spectacular, its decline in seats fragmentation in this election is also near the record. The fall from 5.02 in 2022 to 2.5 in 2026, a change of –2.52, is tied or exceeded (again, among countries with at least ten elections through 2000) only by two countries: Israel 2003 again (–2.52) and Netherlands 1977 (–2.71).

As discussed in greater length at the 2022 planting, Costa Rica’s seat product of 464 is modest for a proportional system, and should be expected, in the absence of any other information, to yield a largest party with around 46% of the seats, an effective number of seat-winning parties (N_S) of 2.78, and N_V=3.16. So this election has resulted in a party system somewhat less fragmented in terms of seats (largest party with 54.4% and N_S=2.50) but still very slightly more fragmented than expected by votes (3.42). Compared to the hyper-fragmentation of the three prior elections, Costa Rica 2026 is pretty much a “normal” election for its electoral system.

____

2002 was also Costa Rica’s first presidential election to require a runoff. Another runoff would not be required till 2014, and then would be required in each of the next two elections. ↩︎
The earlier post actually says 8.3, but this appears to have been in error, albeit only very marginally. ↩︎
This was in the first of two elections in 2012, and the value has come down in subsequent elections. ↩︎
In 2023 it came all the way down to 10.25. ↩︎
The 2025 election produced a record low (for the Netherlands) seat share for the largest party, but due to the shape of the rest of the field, the effective number of parties was not as high as in 2021, although it was again over 8. ↩︎
Note that this seat product (assembly size, times mean district magnitude) is only a little higher than that of the US (435) and significantly below that of the UK (650). It is thus a relatively “majoritarian” system by the standards of “proportional representation’; in other words, a seat majority is not a surprise under this electoral system, even if we would expect the norm to be somewhat short of a majority. ↩︎
The sample mean is 1.50, although we should probably look at the ratio for concurrent elections, which is a bit lower, at 1.39. In this election, the effective number of presidential candidates was 2.85; the ratio of effective numbers between assembly and presidential votes is 1.2. ↩︎
It is generally agreed by scholars of this period that direct election of the PM gave incentive to greater assembly-vote fragmentation, because after voting for their preferred leader of the country, voters felt free to choose an ideological or sectoral small party. While this much could be true of a presidential system, too, the crucial difference is that the executive in Israel remained dependent on assembly confidence and thus the leverage of the small parties remained significant, increasing incentive to split the ticket. For the 2003 election there was also an increase in the legal threshold (from 1.5% to 2%). ↩︎

Thailand 2026

Posted on 18/02/2026 by msshugart

Thailand held a general election on 8 February. The conservative Bhumjaitai Party won a large plurality of seats, although well short of a majority. Thailand uses a mixed-member majoritarian (MMM) electoral system in which there are 400 constituency seats and 100 nationwide party-list seats, with the latter seats simply added on to any given party’s constituency total (“parallel” allocation as distinct from the compensatory allocation of such seats under MMP).

Bhumjaitai won 174 of the 400 nominal-tier seats and 193 overall, despite only 18.1% of the party-list votes. It did substantially better in the nominal votes, with 29.9% of the votes, but still its overall performance, amounting to 193 total seats is a substantial overrepresentation. That is, the parties’ total seats are 38.6%, a percentage that is double its percentage of the party vote. Those 174 seats come to 43.5% of the nominal-tier seats, and this is the bonus that MMM is not able (or intended) to compensate for, but only to soften through the allocation of those list seats.

Notably, Bhumjaitai was only the second largest party in the party-list vote, although it was first in nominal votes. The People’s Party won 29.7% of the list votes but just 23.6% of the nominal votes. More importantly, given MMM, it won only 87 seats in the nominal tier. To that is added 31 list seats for a total of 118.

The different outcomes of the two tiers are quite remarkable as we go farther down the table of results. The third largest party was straightforward enough. In both list and nominal votes that was the Pheu Thai Party, with 15.6% of the list vote and 17.4% of the nominal vote. This resulted in 16 list seats to add to its 58 nominal seats, which was also the third largest total. But in fourth place in total seats is the Kla Tham Party with 58. Of these, 56 were won in constituencies, on 11.5% of the votes but only 2 from the list, for which it won 1.8% of the votes (8th highest). The Democrat Party was fourth in list votes and seats (11.1% and 12), but won only 10 constituencies (on 6% of the nominal vote). Several other parties won at least 1 list seat–there is no legal threshold–and five other parties won nominal seats. (The smallest party to get a list seat was Palang Pracharath which had just 0.4% of the list votes; it also won 4 nominal seats!)¹

All in all, this is a result that almost looks like two different electoral systems–which, given how MMM works, it more or less was–and even like two different elections. MMM rewards a party that can win many constituency seats even if it is not the most popular party nationwide. And that perfectly describes Bhumjaitai, a rather distant second on party votes but strong at the constituency level. To a lesser degree it also describes Kla Tham.²

I know far too little about Thai parties or the country’s current politics more generally to know what to expect for government formation. This was the third election under the 2017 constitution, implemented following the 2014 military coup. Initially the new constitution was supposed to implement MMP, but after one election that actually was MMP,³ the system actually in use has reverted to MMM, more or less identical to what was used for several elections starting in 2001.

_________

Palang Pracharath’s total nominal votes was 1.4%. An even smaller party managed a single constituency win: New Opportunity Party, which had 0.5% of the total nationwide nominal vote and 0.2% of the list vote. ↩︎
And even the little Palang Pracharath. ↩︎
In 2019, with a single fused vote; that is, the votes used to determine the overall seat entitlement (and hence how many compensatory list seats, if any, a party wins) were simply summed from all a party’s nominal votes. ↩︎

Japan 2026: A landslide enabled by the electoral system, but tempered by the nomination strategy

Posted on 13/02/2026 by msshugart

Japan held a snap general election on 8 February. The result was, by any definition, a landslide, especially in terms of seats. The Liberal Democratic Party (LDP) won 316 of the 465 seats. That’s 67.96%, which means the LDP governing majority now will have the votes to override any objections on legislation from the House of Councillors (the upper house, in which the LDP does not have a majority). The LDP won the vote by a wide margin, as well, albeit not quite a majority. Under Japan’s mixed-member majoritarian (MMM) system, there are two votes–nominal (single-seat district) and list–and the two components of the system are allocated in “parallel” meaning the list seats a party wins are simply added to the districts it has won. There is no compensatory mechanism, as there is in MMP systems. The LDP won 49.1% of the nominal vote and only 36.7% of the party-list vote. Thus the electoral system really made this landslide: the total seat share is 1.39 times the nominal-vote share and a whopping 1.85 times the list-vote share. However, the seat landslide could have been even bigger. The LDP actually had to forfeit 14 seats that it could have won, due to having insufficient candidates on its party lists.

The impact of MMM is clearly seen in the fact that the LDP’s 49.1% of the nominal vote translated into winning 249 of the 289 seats, which is 86.2%! There were 31 prefectures in which the LDP swept all the single-seat districts. (Kudos to Nippon.com for actually using the term “single-seat constituencies” instead of “single-member” districts/constituencies!) To this base number of 249 seats was added the 67 list seats (out of 176 total) awarded to the LDP, which would have been 81 had it presented more list candidates. The presence of the list component of the system thus substantially reduced the LDP’s seat share from what it was in the single-seat districts–after all, 68% (or even 71% if it had presented enough candidates) is a good deal less than 86%. But this is not “compensation” in electoral-system sense because, were the Japanese system MMP, the LDP would have won no list seats (at least if the compensation were nationwide) and thus would have had 249 of the total 465 seats, or a mere 53.5%.

The LDP’s deficit of (list) candidates results from the combination of the massive landslide and the tendency to nominate many candidates both in a district and on the list. Or, rather, lists. That is, the “problem” of not having enough list seats to claim all its potential seats is also partly a product of the lists being regionalized. Instead of a single nationwide list, as in many mixed-member systems, Japan elects the list component from eleven regional districts. In four of those it ran out of candidates on the list. Forfeiting seats in this manner has occurred before, but this was “believed to be the first time for a political party in Japan to have faced a shortage of 10 or more potential lawmakers” (Nippon.com). In other words, even the LDP itself did not expect to win this big. Moreover, the extremely short campaign period may have inhibited recruiting additional candidates.

The runner-up party in terms of seats won from single-seat districts (SSDs) was the Japan Innovation Party (JIP), which won 20 seats despite being only the fifth most voted party in nominal votes (6.6%; it was fourth among party lists with 8.6%). As can be the case with single-seat districts, the JIP benefits from being regionally concentrated. By contrast, the second party in votes (of either kind), the Centrist Reform Alliance (CRA) won only 7 SSDs for its 21.6% of the nominal vote. The CRA also won 18.2% of the party-list votes which netted it 42 of the list seats (23.9%). The JIP had been supporting the minority LDP cabinet that resulted from the 2024 election. (As I noted at the time, given Japan’s electoral system, a leading party or alliance being short of a majority of seats actually should be “normal” even though the system is mixed-member majoritarian.) The LDP had sought the JIP’s support following the loss of support from the Komeito, which had been the LDP’s pre-election alliance partner over several elections. However, importantly, the agreement with the JIP, unlike the old one with Komeito, did not extend to electoral coordination. Prior to this election, Komeito was absorbed into the new CRA in a merger with the Constitutional Democratic Party (CDP). The CRA was actually third in nominal seats won in this election, despite being second in votes. The Democratic Party for the People won 8 SSDs on just 7.5% of the nominal votes (it also won 20 list seats on 9.7% of list votes).

It must be emphasized that this landslide really was made possible by the electoral system and opposition fragmentation more so than LDP popularity. The list vote for the LDP was not unusually high. Its 36.7% was a big rebound from 2024, when it was only 26.7%, but not so far ahead of 2021, when the LDP’s list votes were 34.7% and it won 72 of the 176 total list seats and 259 overall.

In terms of nominal votes, even 49.1% is not extraordinary. For instance, the LDP’s nominal vote percentage in 2021 was 48.1%. Yet in that election it won 187 of the 289 SSDs (64.7%), a far cry from the 249 it won in this most recent election.

A key to this bigger win in terms of seats is that the most popular opposition party fell well below half the leading party’s votes. Under MMM, with its priority in overall allocation to whichever party can perform best in SSDs, having a big lead can be more important in terms of the seat payoff than the party’s absolute level of support. (The CDP had just under 30% of nominal votes in 2021.) The 2026 result is perhaps most similar to those of 2017 and 2014, the latter having been another election that was called very early, but even then the LDP’s dominance of the nominal tier contests was not as great.

It was thus a good time for the Prime Minister Sanae Takaichi to deploy her prerogative to dissolve the House of Representatives and call a snap election–a power Japanese PMs have used frequently to surprise the opposition. The combination of the MMM electoral system’s disproportionality and the opposition being in disarray made a huge landslide possible, even if it could have been a little bigger if only the LDP had presented a few more list candidates.

Barbados 2026: Third straight complete sweep

Posted on 13/02/2026 by msshugart

Barbados held a general election on 11 February. For the third straight election, the Barbados Labour Party (BLP) won all 30 seats. It managed its clean sweep on 69.8% of the votes, which is slightly up on 2022 (69.3%) but still below the first in this sequence, 2018, when the party won 73.5% in defeating the then-incumbent Democratic Labour Party (DLP). Barbados uses single-seat plurality, also known as first past the post (FPTP). Sweeps under FPTP in small assemblies do happen from time to time. For national-level elections, this is my complete list:

country	year	assembly size
Barbados	2018	30
Barbados	2022	30
Grenada	1999	15
Grenada	2013	15
Grenada	2018	15
St. Vincecnt and the Grenadines	1951	8
St. Vincecnt and the Grenadines	1989	15
Trinidad & Tobago	1971	36

Out of more than a thousand national elections in democracies in my dataset, I have eight cases of a party winning 100% of the seats. Note that Barbados now will tie Grenada for the most in one country. But Grenada has an assembly half the size of Barbados. The smaller the assembly, the more likely such an outcome will tend to be. From the seat product model, we would expect the largest party, in the absence of any other information, to have 65.4% of 30 seats (19.6) or 71.3% (10.7) of 15 seats, as in Grenada. So a lopsided result is to be expected as long as these countries continue to use FPTP in such small assemblies. A complete sweep is just an extreme case of lopsidedness.

Moreover, as noted a while back, the standard deviation of the ratio of actual largest party share to SPM-expected share tends to be greater with a smaller seat product. (When all seats are elected from single-seat districts, the seat product is equivalent to the assembly size.) Even granted the greater variation expected (and observed) from low assembly size, a ratio of observed to expected largest share that is great than 1.5 is at almost the 99th percentile of all elections held with seat products of 40 or smaller. A complete sweep of 30 seats is a ratio equivalent to 1.53. Thus a sweep should not be especially common, and indeed 8 (now 9, or more if I have missed some) out of 108 elections in this range of seat product is certainly not common. Yet it has now happened in Barbados for three elections in a row.

Such a result is hardly good for democracy, even if the winning party is obviously popular. The most straightforward thing Barbados could do to minimize the chance of such an outcome in the future would be to increase its assembly size. At 30 seats, this is less than half the expectation under the cube root law for a country with a population of around 280,000. With 65 seats (the cube root of the population) and FPTP, sweeps presumably would be harder to come by. At least in theory. However, Barbados may have such a uniform distribution of party support that even a doubling (or more) of the number of single-seat districts might not solve the problem. The closest individual district result in the 2026 election was St. Lucy with a result of 50.2–42.4%. I counted five others with the winner on 53% or lower. The second largest party nationwide should be expected to win a few seats in a larger assembly, but this would still be a small effect when you simply do not have a lot of regional variation.

Even with a 65-seat assembly, given FPTP, the largest party would be expected, in the absence of any other information, to win 59% of the seats (39). From such an expected largest party size, the best guess of vote percentage would be 50.8%. From the 30-seat assembly, the best guess would be 54.7%. Obviously the BLP has exceeded this greatly over this three-election sequence. Thus the only solution would be a form of proportional representation (PR) which would also have the obvious advantage of allowing the assembly size to stay well below my suggested–yet not likely sufficient–doubling in size.

Is there a movement for PR in Barbados? It looks like there is, but it may be marginal. A group of smaller parties (of course!) formed a People’s Coalition for Progress, which includes PR in its list of proposals. This party managed 0.89% of the votes, contesting just 12 of the districts. (In one, St. Michael’s West, its candidate got zero votes.) There is a website called Barbados Underground that advocates PR. It makes this observation:

What happens to constituency representation under proportional representation? All constituencies disappear . . . Barbados is much too small to be carved up into this maze of constituencies. A constituency of only 5 000 registered voters enables rum and corn beef politics because at that scale vote buying is a winning strategy.

Over the years, I have heard many different terms for local constituency-oriented politics, but I must admit “rum and corn beef politics” was a new one on me.

(It is worth noting that a larger assembly with FPTP would make “rum and corn beef” even more prevalent, by making the number of voters per district even smaller. This may not be a good thing! There is also some malapportionment, although not to the extent seen in Belize.)

If Barbados were to stick to a 30-seat assembly but elect all 30 in a single nationwide district, the expected largest party size would fall to 42.7% (about 13 seats). There is precedent in the region for such an electoral system: Aruba and Curacao have 21-seat nationwide districts. I have a few of these elections in the dataset, and for those from Aruba I also have the seat share of the second party:

country	year	assembly size	s1	s2_nat
Aruba	2005	21	0.5238096	0.38095
Aruba	2009	21	0.5714286	0.38095
Aruba	2013	21	0.6190476	0.3333
Aruba	2017	21	0.4285714	0.4285714
Curacao	2012	21	0.2380952	.
Curacao	2021	21	0.4286	.

Of course, there are other models of PR that could be considered in Barbados. I have no concept of regions within Barbados and whether the above quote about no need for constituencies is a reasonable claim. (People presumably do like their rum and corn beef.) One could imagine six districts of five members each, or other possibilities falling within the spectrum between FPTP and single nationwide PR, even without adjusting assembly size. Or an MMP model with the current 30 constituencies and 10–20 or so compensatory seats.

Of course, all the above assumes that complete sweeps are a problem. Maybe they are not! I doubt it, though. Democracy should have one or more opposition parties with seats in the assembly.

It is also possible that this will right itself, and future elections will be more competitive. As recently as 2013, the election was close. The DLP won 16 seats and the BLP 14, with votes splitting 51.3–48.2%. The election before that (2008) saw the DLP win 20–10 on seats (almost exactly the seat-product expectation) and 52.6–47.2% on votes. However, even leaving out the recent sweeps, the mean seat share across ten Barbadian elections has been 71.9%. So my (totally unsolicited) advice would be minimally to increase assembly size significantly, but to think more ambitiously about electoral reform, which as noted, would not require an assembly size increase or would work with only a small increase. Complete sweeps are an outcome a democratic electoral system should be designed to make unlikely.

Chill and early bloom, 2026 (and it is 2/2 Tu Bi-Shvat!)

Posted on 02/02/2026 by msshugart

We have accumulated a good amount of chill already as February begins. In fact, so much that blooms are already getting underway. Today was 2/2, Tu Bi-Shvat 5786, or around the time the almond trees should begin blooming, per tradition, in the Land of Israel. Here in the Sacramento Valley of California, the climate is quite similar, and so almonds do tend to have their first blooms around now. In fact, of my two almond trees, one very technically has made it “on time.” At first glance, it does not look like much. But there is actually one bud unfurling as of this afternoon. (The local commercial almond groves always bloom a good deal later than my two seedling trees, which produce worthless nuts.)

This year’s chill-accumulation pattern has been unusual. As the graph below, from the nearest CIMIS station, shows, as the calendar turned to December, our total chill hours received was well below the three immediately prior years. Then all of a sudden it ratcheted up rapidly. This station had recorded 110 chill hours–using the definition of hours between 32 and 45F–compared to 242 at the start of December 2022, and 175 and 169 in the two years between then and now. As of the 16th of December, the 2025-26 chill count stood at 428. Amazingly, it had quadrupled in just over two weeks! Of the four most recent winters, only in 2022 had it reached 400 (barely) while the other two years it was still under 300 at that point. (My location gets a somewhat greater chill accumulation in a typical winter than this CIMIS site. In the linked post I note that I prefer the “Utah model,” which as you can see below says this site is already over 1,100 hours or “units.” However, most nurseries and other professionals continue use “below 45” or “32 to 45.”)

That steep uptick is what more than two weeks of essentially never seeing the sun can do for you (in addition to being quite depressing)! We had the most intense and prolonged Tule fog/stratus that I have experienced in a dozen winters in this valley. For nearly that entire period, the temperature was perpetually in the range of 40 to 47F, which happens to be almost exactly the range of prime chill accumulation for stone fruit trees.

In the second half of the month, chill basically stood still. This was during our one significant rainy period of the winter thus far. With storm clouds and abundant subtropical moisture, lows seldom got below 50, yet highs were also in the 50s or briefly low 60s. In this range, there is no positive chill, but also no “negative chill.” (As I have written about before, I am skeptical that you actually need to subtract chill hours during winter warm spells, but warm temperatures certainly hasten bud break, even if a tree’s chill has not been fully achieved. During this time, there was nothing hastening growth of anything, aside from weeds.)

As is also notable from the graph, January also was good for chill. The rate of increase was much less than steep than in the first half of December, but it was still good. This period saw most nights get to the mid-to-high 30s, and a few mornings right around freezing. It also saw many highs in the 60s. In January, per the CIMIS records, we were averaging about 8.4 chill hours a day (via the 32–45 method). With the first seven days of the month still being mild and recording zero additional chill hours, the chilly period of January was actually just 23 days during which we added 259 hours, an average of 11.3 per day. Not as good as the 22-day period of Tule fog that actually started on 25 November and continued through 15 December when we got 415 chill hours, an average of 18.9 per day. But still pretty good.

As the graph above shows, we usually can count on additional chill during February and even March, although not necessarily at as steep a rate as we have been having (though it did continue at about that rate in 2023). So the final count could go well over 1,000. On the other hand, most trees will be indifferent to February/March chill this year. First of all, I have few varieties that need more than about 800 hours, and most of what I have planted is listed as being in the 500–700 range of chill requirement. In fact, I am somewhat surprised that nothing began to bloom until the 30th of January when the Flavorella plumcot had its first bloom. Here is is, pictured today. This view shows buds about to burst, although not the few that are already open. Look closely, there is a bee trying hard on a bud that is not yet open. (The bee is out of focus; I was using only the iPhone camera, and it struggles with this sort of shot.)

Flavorella is almost always my first stone fruit to bloom–usually ahead of the almonds and only occasionally beaten a day or two by either Splash or Emerald Drop pluot. This is the earliest bloom I have seen on any stone fruit in the dozen years I have been here, although Flavorella has had its first blooms in the first week of February before.

I really expected several varieties to be blooming before now. As early as the 9th of January, I had visible bud swelling on several trees, including the Candy Heart pluerry, pictured below. But so far, aside from several Flavorella buds and that one almond bud, nothing is open yet. Apparently the return of chill around the middle of the month slowed down the progression of the buds, at least for a while.

That is good, of course, as blooms in January would not be desirable, precisely because we could still have weeks ahead of cold temperatures, even freezes. This remains a risk now even though the orchard as a whole has held off from a January bloom. We have had serious freezes as late as the third week of February, and if that happens this year, it could be bad. There are long-range forecasts of a possible cold snap by the middle of the month. I am hoping it will not be too deep a snap.

It seems spring is upon us, even if winter is not really behind us. If the next months offer reasonably mild weather, we could have a good fruit harvest. But it is way too early to assess that. In the meantime, I will enjoy the show as more and more trees come into bloom.

Finally, I leave this planting with a longer view from out near the orchard on a beautiful sunny Tu Bi-Shvat.

Comment moderation is (partially) on

Posted on 28/01/2026 by msshugart

I always welcome new commenters! However, I do expect them to maintain compliance with a very basic comment policy (as stated at the About page).

I generally do not moderate comments (although sometimes the blog software holds back a comment for one reason or another). For the time being, at least, I am turning on moderation for first-time commenters.

I would rather not do this, but I have had a few first-timers here just recently who were rude, engaged in ad hominem attacks on me, or otherwise were inappropriate in their comment. So I will screen new comments from first-timers and so how that works.

If a comment gets held up in moderation (for whatever reason), I will not always know for a day or two. But any acceptable comment will be manually cleared. As I say at the “About” page, I have sole authority to decide when a comment can be a part of this blog. Over the years, I have rarely refused permission to comment. But there are now and then those who feel they need to make a statement that they are capable of making only by being rude. Good riddance to that sort of person!

It is never my intention to prevent reasonable debate. I value the community that has been built up over the years at F&V. This decision is only to protect what we have that I cherish–that is, protecting the virtual orchard from noxious weeds.

(Comments are off on this post, but not on the rest, aside from a very few where they were already turned off.)

What about the Senate?

Posted on 27/01/2026 by msshugart

What do advocates for proportional representation (PR) in the US House say about the Senate? My general sense is they say nothing. Obviously, the current arrangement in the Senate is not conducive to PR, given that each state elects its two senators at different times (or in separate contests even on those occasions when both seats are at stake due to a special election for one of them). This would be the obvious justification for advocates leaving the Senate aside. So my question is really, how do advocates see a PR-elected House working with a Senate that remains fundamentally unreformed?

I bring this up sometimes in conversations online and elsewhere, but I do not think I have ever seen an adequate response. I also do not think I have ever articulated a position on this at the blog. I am very much on the record as favoring PR for the House. However, the Senate is a co-equal body so having it unreformed while the House is elected by PR may have many undesirable and unanticipated consequences.

If you are a PR advocate for the US House, how do you see it working with the Senate? For purposes of the conversation I hope to spark here, let’s set down the parameters that you can’t have more than two senators per state, in any state. We are stuck with equal representation per state, and the constitution itself would have to be amended to increase the number per state to one more friendly to PR.

Obviously, a variant of this question could be asked about PR and presidentialism. All the more so because of the electoral college, as currently operating, for the presidency. This is also a potential obstacle to having a functional PR system for the House. This is a separate–but very important–question. It is at least one that has been addressed. (See for example, the study by Drutman and Mainwaring.) I believe hardly anyone has addressed the Senate in this context. That seems like a problem, so let’s think about how to resolve it.

What is the theory of regime change via air power?

Posted on 13/01/2026 by msshugart

As I type this, it looks increasingly likely that President Trump will order military strikes on the Iranian regime. Hence the question in the title. I really am curious about this, and to answer my question you do not need to favor action or even believe the “theory” you articulate.

I would not be asking were I not skeptical there is any such credible theory. I have seen various articles and social-media posts laying out possible options. (One example: Yonah Jeremy Bob in The Jerusalem Post today; there are many others, but I have not saved many of them other than a somewhat outlandish “most realistic scenario” posted a few days ago on an account called Iran Spectator.)

The problem is, you can’t dismantle a repressive apparatus via airstrikes. (Or can you? That is part of the question, I guess.) There is risk of the Iranian leaders escalating attacks on protesters. There is some (unknown, but be prepared) risk they fire missile barrages at Israel again, or at US bases. I lack the knowledge to game this out. I just hope US leaders, including the President, know what they are doing. On that front, I want to be optimistic, but I struggle to be.

So what is the theory they might be working with, and is it in some way plausible? We may be finding out soon. 2026 is only in its 13th day, and already it has been rather too interesting.

Japan: Assembly-size reduction fails, return to SNTV being discussed

Posted on 08/01/2026 by msshugart

A proposal to reduce the size of the Japanese House of Representatives by 10% failed to pass in the autumn session. Meanwhile, there is discussion of reverting to Japan’s former electoral system of single non-transferable vote (SNTV). Via Japan Times, 2 January.

The assembly size reduction was part of the coalition agreement between the Liberal Democratic Party (LDP) and the Japan Innovation Party (JIP) that was signed last year. As an earlier (26 October 2025) Japan Times article noted that “Small parties that rely on proportional representation seats are on high alert,” because the plan would remove about 50 party-list seats allocated by proportional representation in Japan’s current mixed-member majoritarian system.¹ Yoshihiko Noda, leader of the main opposition Constitutional Democratic Party of Japan, has called for the cut to include a reduction of the number of single-seat districts. As Japan Times notes, this would complicate the electoral “coordination” of the LDP.

The recent (2 Jan.) article on possible electoral system change does not refer to any specific proposal to bring back SNTV, but indicates there is “growing debate in parliament” about doing so. The justification given is:

Advocates for scrapping the current model say a multiseat district system would better allow parliament to reflect a diverse range of views and spur more voter interest by helping to cut down on “dead votes,” or ballots cast for candidates who end up losing their single-seat district race.

This is a specious line of argument. If these were really the goals, there would be better ways to achieve them. The only advantage to SNTV is it is familiar, given its use to elect the House of Representatives for decades prior to the mid-90s. However, better ways to achieve the stated goals would be either to convert the mixed member system to the proportional (MMP) type, or adopt list PR. Given the importance of personal votes in the LDP, open-list PR would be a sensible choice. In fact, half of the House of Councillors is already elected by open-list PR.

It actually gets worse, in that the recent JT article indicates some of the proposals circulating around would be for the limited vote, not SNTV. For instance, allowing the voter to vote for two candidates in a three-seat district. SNTV is, of course, a special case of limited vote, in the sense of limiting the number of votes to not simply less than the number elected in the district (the district magnitude, or M), but all the way down to one. Both have the basic feature of being pure candidate-based systems where the top M candidates win, regardless of party affiliation. By contrast, open-list PR would first pool the candidates’ votes at the level of the party list, before determining how many seats each party would win.

The Democratic Party for the People (DPP) is said to believe the limited vote “would avoid a concentration of power in a single party and foster a multiparty system.” This is seriously misguided. Most voters could be expected to give their full allotment of votes to candidates of the same party, and thus the limited vote, giving multiple votes per voter, is more favorable to the largest party than SNTV is. (Basic electoral-system theory!)

Readers interested in a discussion of SNTV may want to consult the recent post that started off being about Kyrgyzstan’s adoption of such system, but morphed into a very interesting discussion in the comments about the place of SNTV in the family of electoral systems (and related topics).

Thanks to those who called the first-linked JT story to my attention–Yibo in another thread here and my colleague, Ellis Krauss.

One of the articles says it would be a reduction of 25 constituency seats and 20 list seats. It is not clear which proposal actually was voted on. ↩︎

Ranked choice voting at the county level in California: Santa Clara’s “unique circumstances”

Posted on 04/01/2026 by msshugart

This post is a follow-on to the previous one about a low-turnout runoff in Santa Clara County, California. Could the county change the election of its officers to ranked-choice voting (via the alternative vote a.k.a. instant runoff)? Yes, it could. However, it is the only county that currently has this authority. Further, it appears it could switch to STV for its Board of Supervisors.

AB 1227 of 2023 was passed by the state legislature and signed by the governor and explicitly grants to Santa Clara County the option to adopt ranked-choice voting (RCV) for any or all elected county officers. It allows that either the county’s Board of Supervisors or a county initiative may enact RCV.

The legislation further says, “The ordinance [to adopt RCV] shall specify which county officers shall be elected by this method and whether they shall be elected at large or by or from district, as applicable.” The reference to “at large” surely enables the country to elect the Board or other multi-person bodies in the county by single transferable vote (STV), sometimes called (in the US) proportional ranked-choice voting (PRCV). The first section of the enacted state law–the “findings” section–references cities in California that have adopted STV, such as Albany and Palm Desert, although it simply refers to their systems as “RCV” without the important detail. (Palm Desert now has two districts, one of which is a single seat using RCV/alternative vote, and the other of which elects four by STV/PRCV.)

Interestingly, the law applies only to Santa Clara County. Moreover, its third of three sections reads, in full: “The Legislature finds and declares that a special statute is necessary and that a general statute cannot be made applicable within the meaning of Section 16 of Article IV of the California Constitution because of the unique circumstances relating to the County of Santa Clara’s interest in having the option to use ranked choice voting in its elections.” (The cited section states, “A local or special statute is invalid in any case if a general statute can be made applicable.”)

Therefore, we might say that we have here a case “electoral system pork barrel”! Okay it is not a project, which is what “pork barrel” implies, but a measure regarding electoral systems that provides a benefit only to the geographical region represented by its sponsors. The originally introduced bill did not have that third section, although it did already state its applicability was only to the one county. The introducing member was Assembly Member Evan Low, who at the time represented Assembly District 26, encompassing parts of the Silicon Valley, which is to say a large portion of Santa Clara County. Assembly Member Alex Lee later joined as a co-author; his district encompasses other parts of the county, including part of the city of San Jose.

Curiously, the final chaptered amendment to the state code says that an election under RCV in Santa Clara would have to take place at the same time as the statewide “primary” (i.e., the first round under California’s two-round not-a-primary electoral system used for state and federal offices). It is very strange that it would not be held at the November “general” election. This provision changed through the bill’s legislative process, and I will address the apparent reasons for this below. The initial version did not specify the timing of a potential RCV election in the county. The first amended version in the state Assembly added this tortured phrase: “If a proposal would elect an officer by ranked choice voting through a single election with no possibility of a runoff, the bill would authorize the county to hold that election at the first statewide general election following the statewide primary election at which the election otherwise would be held in accordance with existing law.” An amended version in the state Senate said in somewhat more straightforward language that the county could hold an RCV election at either the “primary” or the “general” while adding that third section quoted above.

A further round of amendments in the Senate saw two members of that chamber join as further coauthors, Josh Becker and David Cortese, both of whom represent–you guessed it–parts of Santa Clara County. However, before its final passage the language to require that any such RCV election in the county be held at the “primary” was included.

The bill’s Assembly Floor Analysis dated 9/05/2023 (available at LegInfo) indicates that the jurisdictions within the state that have so far adopted RCV are all charter cities, but that it is not clear under current law that charter counties would have this authority, because counties are generally given narrower discretion in their election ordinances.

The county itself had determined that it did not have the authority absent a change in state law. Thus what we have here is a change in state law for this one county for the simple reason that only this county had applied for authorization. It did so in response to a countrywide ballot measure that had indicated the county would adopt RCV if it became possible within the capacity of the vote tabulation machinery used by the county. That measure passed in 1998, but only in December 2022 did the county Registrar of Voters declare that it now had the relevant machinery.

It is not clear from the immediately available resources whether there was any debate on making this legislation a blanket permission for counties to adopt RCV. All versions of the bill as it worked its way through the legislature explicitly were only about Santa Clara County.

As for the question of the date on which an RCV election could take place, it seems this is due to the potential incompatibility with state law if it were set to take place in November. This is implied by the report on the Assembly Third Reading, dated April 24, 2023 (also available at LegInfo). It notes that elections for county officers are required to be held at the statewide “primary” with a second round in November only if there was no majority in the first round. Therefore, a county holding no election for such an office at the first round would be in potential breach of state law, and evidently legal counsel said that even the third section added to the bill would not remedy that problem. This is unfortunate, of course. RCV (in the form of the alternative vote/AV) is by definition a one-round solution to obtaining a definitive result that might otherwise require two rounds to obtain a majority. So logically it should be held at the end of what would otherwise be a two-round process, not at the beginning. For sure turnout is higher in November than at the first round. Not for the first time, logic gets tripped up by the law!

The bill ultimately passed the Assembly 67-0 with 13 abstentions and the Senate 31-7 with 2 abstentions. I do not know if there is currently a strong prospect that the county will go ahead and be a pioneer in using RCV for any of its county offices, but possibly that low-turnout Assessor runoff special election will reenergize the debate.