On a proportional Westminster system

The main idea: Proportional Westminster government by way of either parallel voting or random ballot.  

0. Posts on this blog are ranked in decreasing order of likeability to myself. This entry was originally posted on 22.09.2022, and the current version may have been updated several times from its original form.

 

1 Business as usual

1.1 A rather obvious and low-risk way of making a Westminster design more proportional is to elect the vast majority of the House through single-winner ridings, and a small minority (a fifth to a quarter) from a nation-wide party-list. Such a small top-up is more than enough to ensure good (not perfect) proportionality. This would be enacted by issuing each voter two linked ballots, one for their riding (with the whole country allocated into ridings) and a nation-wide ballot listing parties only.

1.2 Those competing in the single-winner districts would not be allowed to also feature on their party’s closed lists and, more crucially, you could only elevate a minister from the ranks of the single-district winners, not all House members at large.

1.21 Those who otherwise tolerate bicameralism could also consider having each vote require a double majority of the single member winners only and the entire House as a whole. I myself am not too keen on such tricks though, as I like my political accountability clear.

1.3 This is now still a recognizably Westminster-type of setup, albeit one much more proportional.

1.4 Now, the setup is not a Mixed Member Proportional system, which causes all sort of issues with decoy lists, but a parallel system that still creates national-wide proportionality.

1.5 You calculate a quota of support needed to elect one member of the House at large (not just those competing in single districts), and remove that much from the party-list votes of those whose candidate won in the single districts (remember that the two votes are linked). You use the de-weighted votes to elect the members of the party-list seats all at once as per your favourite proportional rule, ignoring who won the single-winner districts. Thus, the two systems combine to allow for near proportionality.

1.6 I say near proportionality since it can happen that the winner of some single district received less than a full quota of support, in which case you just de-weight those voters’ ballots to zero in the proportional part, instead of going negative. It’s all right, it’s a proportional-enough system, not a fully proportional one.

1.7 Note how the design allows you to use any sort of vote as your riding vote, not necessarily the shonky vote-for-one design. You can, for example, easily have the riding vote be approval voting.

2 The random ballot

2.1 But if you want true proportionality - and not just in terms of party affiliation, but in terms of gender, ethnicity and all other sorts of characteristics of the candidates as reflected by the votes they get - you need the system that elects each riding’s winner through a random vote, with no top-up of other members except these.

2.2 Whilst this is obvious, far less obvious is how to implement the selection of a random ballot in a safe way, so here’s my take.

2.3 After polls close, the watchers assemble and take each box to perform the following procedure: dump all votes on the table and (assuming two major parties), one party’s representative splits the bundle into two halves (roughly), and the other party’s representative selects one of the bundles.

2.4 On this bundle, the procedure is repeated, until we get three or four such halvings (the number of halvings has to be exactly the same for all boxes and all locations, but it’d be either three or four). At this point, the few votes remaining are boxed-up in and sent into a centralised counting location, with each riding having one of these.

2.5 There, the process is repeated (with the candidates serving in the role of the party representatives) until you are left with one vote only, which determines you winner. If this vote happens to be invalid, go back to the other half you dismissed, and so on.

2.6 Whilst the design of the vote and its folding may require some rethinking such as to prevent the selection from being read with the vote on the table, the system is otherwise ready to be implemented today as it requires no further infrastructure from what you already need to count an election.

2.7 Not tamper-proof but at least as safe as any hand count anywhere on earth, and far, far safer than any electronic count at a comparable cost in terms of time. You can get the results within two hours at most.

2.8 The above assumes you are happy to randomise the "vote for one" ballot, which you shouldn't be, so here's how to randomise the far superior approval voting: voting on a (mechanical) machine approval style (vote for as many candidates as you like, but one vote each at most), with the machine just printing n vote-for-one ballots where n is the number of candidates you approved off. You check the prints visually and, if you are happy, put them in the box (again, design required to make sure these are not easily legible, maybe ink that fades after minutes of exposure idk). Note how converting an approval vote into individual vote-for-ones retains the key advantage of approval voting, where approving of an additional candidates increases his chances but decreases the chances of your other approvals.