A common divisor method for allocation of seats to lists based on their votes in PR systems or to jurisdictions based on their population is named for the Belgian who “invented” the method in Europe.1 But what was the Belgian’s name? Victor d’Hondt? Victor D’Hondt?
I learned d’Hondt. And this seems to be how most of the political science literature on electoral systems spells the name. It is also how the most recent book on the quantitative analysis of electoral systems spells it: Rein Taagepera, Predicting Party Sizes (Oxford, 2007).2 However, the glossary of the edited volume by Michael Gallagher and Paul Mitchell, The Politics of Electoral Systems (Oxford, 2005), says, in part and at p. 632:
Devised by Belgian law professor Victor D’Hondt (1841–1901). Often spelled ‘d’Hondt’ but correctly spelled with a capital D (see any Belgian library catalogue, including that of his former university, the University of Ghent)
That is pretty persuasive (all the more so because the Gallagher/Mitchell volume contains a chapter on Belgium, authored by a Belgian specialist in electoral systems).
So, I am making the transition to D’Hondt3 It is hard, however, as I have spent years “correcting” students or authors of papers I was reviewing when they wrote it as D’Hondt rather than d’Hondt.
It would be so much easier to just refer to it by the name of the man who actually invented an identical rule, in a proposal before Congress in 1792.4 There is, as far as I know, no controversy over how to spell the name of Thomas Jefferson.5
By the way, Wikipedia spells it D’Hondt.
Next up: The great pronunciation controversy. Don’t mis-pronounced it!
_________
- The rule divides the shares (of votes won by parties or populations of jurisdictions) by successive divisors starting with 1 and increasing each by 1 (i.e., 1, 2, 3,…). It then allocates seats according to the resulting quotients. [↩]
- Of course, it was from Taagepera that I learned how to spell the name of the common divisor formula. But, as I noted, he is hardly alone in using that spelling. [↩]
- And wishing the US Democratic Party would do as well. Though if they prefer to use d’Hondt, that would likewise be an improvement. [↩]
- Some other common ‘PR’ formulas also had earlier US inventors for apportionment of House seats among the several states: Sainte-Laguë/Webster and Hare/Hamilton (a quota-and-largest-remainders rule, rather than a divisor method). A very interesting point in Taagepera’s book, at pp. 32-3 (citing Colomer 2004:44), is that d’Hondt is the “remainderless quota” or “sufficient quota”; i.e., the quota for a given votes distribution that is sufficient to allocate all seats without any recourse to remainders. All the more reason why it is arguably the more desirable ‘PR’ method. As Taagepera puts it: “It is at the crossroads of quota and divisor methods.” [↩]
- And, it would be sensible for a US party to use the formula invented by one of the founders of the Democrat-Republican Party! [↩]
It gets funnier: The English Wikipedia notes that D’Hondt is the Flemish and d’Hondt the Dutch (Dutch Dutch?) way of spelling the name. The Dutch Wikipedia makes no notice about this.
It really shouldn’t come as a surprise that a Belgian will give rise to a series of complications, though.
Oh, and I usually pronounce the name as dånt
Seed planted by Jacob Christensen — 01 June 2008 @ 18:46
Just in time to fix it in the page proofs of Voting Amid Violence.
Although I think I “corrected” D’Hondt to d’Hondt in an article manuscript I reviewed the other day for LAPS.
Seed planted by Steven Taylor — 01 June 2008 @ 22:08
Would it be too obscure an allusion to state that I don’t care which dog is the winner of this fight? (small d vs large D)…
Seed planted by Tom Round — 02 June 2008 @ 03:05
I agree with MSS that divisor systems (which can be translated into “all and only remainders over X% of a full quota”, but with a fixed total of seats) are superior to quota and largest remainder systems — especially Hare quota. Divisors avoid the incentive to artificially split lists, and also avoid the “Alabama paradox” (whereby, when the US HR was allocated by Hare quotas and largest remainders – the system entrenched in the Canadian Constitution, BTW – Alabama’s population would have given it 8 seats in a House of 299, but only 7 in a House of 300).
However, while I favour d’Hondt divisors for allocation among “moving targets” (ie, political parties), I prefer St-Lague divisors for allocations among “fixed targets” (ie, electoral districts/ constituencies).
My reasoning being:
1. St-Lague, over a number of apportionments, cancels out some of the bias towards larger units that d’Hondt has. If two or three smaller States/ provinces/ regions together have a combined population equal to that of a single larger State/ province/ region, then over time, their respective numbers of seats will tend to equalise. By contrast, D’Hondt consistently over-represents the single larger unit, while Hare (and some divisor systems, eg Danish or “or part thereof”) consistently over-represent smaller, divided units. St-Lague is the next best thing to allowing preference swapping among regions (“If my home State’s remainder isn’t enough to get us an extra Congressman, I at least want another Midwestern/ Southern/ New Engalnd State to get one”).
2. The above, however, only applies when the competing regions/ constituencies are fixed and unable to either split or merge artificially. Parties, of course, can do both. Used in elections, then, with a large enough number of seats and a low enough supra-quota threshold (or none), st-Lague would in fact enourage artificial splitting. So I support d’Hondt for elections (to the extent that people insist on using party lists rather than STV!), especially if there is a moderate threshold (4-5%) and provision for competing lists to form and alliance and pool their votes (horse-trading in public before polling day, rather than [being seen to do so] in “smoke-filled rooms” afterwards).
Seed planted by Tom Round — 02 June 2008 @ 03:19
There is a different convention on capitalizing a prefix in Dutch surnames :
In Flanders, they are an integral part of the surname and always capitalized: in fact “De†refers to the Dutch “the†(“D’Hondt†comes from “De Hond†meaning “the dogâ€); a little-d “de†would refer to the French particle “de†(of) typical of noble names (e.g. “de Toquevilleâ€).
In the Netherlands, they are not capitalized when the first name is also used: e.g. the name of the Minister of Defence of the Netherlands is spelled “Eimert van Middelkoop†= “Mr Van Middelkoopâ€
Seed planted by Bancki — 02 June 2008 @ 04:06
I think it is reasonable to have different preferred allocation systems for different purposes, and that allocation of seats to parties does not necessarily need to follow the same criteria as other purposes. These other purposes include not only the allocation of seats to electoral districts, but also such tasks as assignment of compensation seats to districts (for which some European countries use different divisors or quotas than they use for the overall distribution of seats among parties).
One clarification: It is not correct that all divisors prevent incentives for splitting, while all quotas create such incentives. As far as I can recall, the advantage from splitting arises from Sainte-Laguë divisors (1, 3, 5, 7…), which may be why they are not used to allocate seats to parties anywhere that I am aware of. Only the modified SL (1.4, 3, 5, 7…) are used. (Regular SL is used for some of the other purposes referred to here.)
As for quotas, the Alabama paradox is specific to simple (Webster/Hare) quotas (votes or population divided by seats) and largest remainders. And that may be why this system is so rarely used. (Many PR systems use simple quotas for a first stage, but then switch to D’Hondt for remainders, which is presumably the same as using D’Hondt all the way.)
It occurs to me that I have no idea what the general tendency is worldwide in the use of methods for allocation of seats to electoral districts.
Of course, for STV, the Droop quota makes the most sense.
Seed planted by MSS — 02 June 2008 @ 11:44
Poor Alabama–it does seem that something is always amiss here when it comes to politics.
Seed planted by Steven Taylor — 02 June 2008 @ 15:04
I stand corrected – using d’Hondt divisors gives a sharp incentive not to split, but other divisors (St-Lague and a fortiori Danish, which divides by 1,, 4, 7… IOW, rounds up every remainder over one-third of the eventual quota) may give such an incentive. Pure St-Lague divisors (1, 3, 5…) would encourage a party with 1.8 quotas to split into three equal blocs, each with 0.6, and thus take 3 seats as against only 2 seats for an undivided rival with “only” 1.9 quotas.
Which is why, as MSS points out, Norway et al modify pure Lague to divide by 1.4, 3, 5… instead, so that a party needs 0.7, not 0.5, of a full quota to get its first seat.
(“Quota”, in relation to divisor systems, is determined ex ante not a priori – ie, it means the largest number that, used as a quota to allocate 1 seat per whole quota or remainder over N% of a quota, will allocate the exact number of seats.)
Seed planted by Tom Round — 02 June 2008 @ 18:00