Proportional Representation: A Study in Methods of Election

by John H. Humphreys

Party A. Party B. Party C. 6,000 4,800 1,900 3,000 2,400 960 2,000 1,600 1,500 1,200 1,200 960 1,000

The eleventh highest number, which constitutes the measuring rod, will be found to be 1000; the largest party obtains six seats, the second party obtains four seats, with a remainder of 800 votes, and the third only one seat, with a remainder of 900 votes. The two smaller parties taken together poll 6700 votes but only obtain five seats, as compared with the six seats obtained by the larger party with 6000 votes; the two remainders of 800 and 900 votes, which together constitute more than a quota, having no influence on the result of the election. Even if, in the allotment of seats, the largest party has a remainder of votes not utilized, yet this remainder necessarily bears a smaller proportion to the total of the votes polled than is the case with a smaller party. Thus the system works steadily in favour of the larger party.

The question of remainders, or votes not utilized in the distribution of seats, is of minor importance when the constituencies return a large number of members. When, for example, as in the city of Brussels, there are twenty-one members to be elected, the votes not utilized bear a small proportion to those that have been taken into account in the allotment of seats. In Belgium, however, there are several constituencies returning as few as three members, and there is naturally a demand that these constituencies should be united so that the method of distribution should yield more accurate results.

If the d'Hondt rule, like every other method of distribution, is open to criticism from the point of view of theoretical perfection, it must be admitted that in practice it yields excellent results. The election at Ghent resulted in the return of six Catholics, three Liberals and two Socialists; it would have been impossible to have allotted the seats more fairly. Under the old non-proportional method the Catholics would have obtained eleven representatives and the Liberals and Socialists none. The immeasurable improvement effected by every true proportional method is apt to be overlooked in the critical examination of the working of these methods in those extreme cases which rarely occur in practice.

The formation of "cartels."

The steady working of the d'Hondt rule in favour of the larger parties has, however, not escaped the attention of advocates of proportional representation. The late Professor Hagenbach-Bischoff has formulated the proposal that parties should be allowed to put forward combined lists, and that in the first allotment of seats the totals of the combined lists should be taken as the basis of distribution. The need of some such provision may be shown by an example used in illustration of the d'Hondt system, at a meeting held under the auspices of the French Proportional Representation League.[8] A constituency with eleven members was taken; four lists, A, B, C, and D, received 6498, 2502, 1499, and 501 votes respectively; the d'Hondt rule made 834 the measuring rod, and gave A seven members, B three, C one, and D none. The question was asked why provision was not made for the transfer of the votes from list D to list C, so that if, for example, these lists were put forward by Radical-Socialists and by Socialists respectively, the parties might obtain the additional seat to which their combined totals entitled them. It will be seen that lists C and D, with a total of 2000 votes (more than twice 834), obtained but one representative, while list A, with 6498 votes, obtained seven representatives.[9]

Professor Hagenbach-Bischoffs proposal, which would meet this difficulty, has not been embodied in the Belgian law, but "cartels" (arrangements for the presentation of a common list) are formed between the Liberals and Socialists so as to lessen their loss of representation due to the working of the d'Hondt rule. The "cartels," however, do not give satisfaction, as experience shows that many Liberals who would vote for a Liberal list decline to vote for a "cartel" of Liberals and Socialists; whilst, on the other hand, extreme Socialists decline to support a Liberal-Socialist coalition. In the Finnish system, however, provision is made for the combination of lists in accordance with Professor Hagenbach-Bischoff's suggestion. Indeed, as the Finnish law forbids any list to contain more than three names, some such provision was necessary in order to allow each separate party to nominate a full list of candidates.

The experience of the Belgian "cartels" would seem to show that, even where party organization and discipline are highly developed, many electors resent the disposal of their votes by a bargain between the organizations concerned. The single transferable vote, by allowing each elector to indicate his second choice in the way in which he himself prefers, would enable smaller parties to obtain their share of representation without involving a preliminary compact between party organizations. A list system seems to establish a rigid division between parties, whilst there is no such corresponding rigid division in the minds of many electors. The model elections conducted by the Proportional Representation Society cannot perhaps be accepted as a conclusive guide to the action of voters at a real election, yet the number of Liberals who, in the last of these elections, gave an effective preference to a representative of the Independent Labour Party, in the person of Mr. Henderson, was very noteworthy. In the Belgian system no such fluidity is possible; the Liberal electors would be shut off from any relation with the supporters of Mr. Henderson, who could figure only upon the Labour Party's list.

The different methods of selecting successful candidates.

It will be seen that the problem of allotting seats to lists has been solved in several different ways. Similarly, different methods have been tried for the purpose of selecting the successful candidates from the respective lists. The instructions to voters vary accordingly. The earlier schemes (and the practice obtains in several Swiss cantons to-day) provided that each elector should have as many votes as there were members to be elected, and that he might distribute (without the privilege of cumulating) his votes over the whole of the candidates nominated, selecting, if he desired, some names from one list, some from another, and some from another. After the number of seats secured by each list had been ascertained those candidates were declared elected who, in the respective lists, had obtained the highest number of individual votes.