Why Gregory surplus vote transfer method is not necessary
- Tom Monto
- Mar 15
- 4 min read
The Gregory Method is often presented as the way to do STV.
I know it has credible defenders, but is it really the way to go?
I think all would admit it is a complicated method compared to Cambrige (city)'s STV, or to the "Exact Method" STV used in Ireland and Malta.
And most if they look would see that all the transfers have little effect on who wins - Transfers seldom change many winners, and sometimes not at all, compared to first round of the count.
Perhaps on average about one in ten owe their election to vote transfers.
Even fewer changes are produced by surplus transfers, which are few in number in any STV district contest.
Four at most in a district with five seats.
I say random transfer of surplus votes is no biggie - if at least the next marked preference is considered, as is done when the Exact method is used.
The "Exact method" where the next marked preference is considered, is 100 percent the same transfer as Gregory concerning how many votes carry what next preference. (of course it maters is al lteh winning candidate's voes are considered or only the "last parcel", but both GM and EM uses either in various situations.
The Exact Method is only random IF the piggybacked lower preferences come into use later, which is not always the case.
There are random methods, and there are really random methods.
Exact method is not really random but it is often lumped in with really random.
Exact Method is not a randomly-shuffled pile of ballots. Under Exact Method, the votes belonging to successful A candidate are sorted into separate piles for next choice.- B pile , C pile, etc.
then from each of those piles the proper number of votes are taken.
secondary preferences are piggybacked in the transfer, and that is where randomness comes in, but those secondary preferences might never be used.
(more info in Montopedia blogs on surplus vote transfers --
Here's big point - the vote transfers seldom make much difference in who wins.
most, sometimes all, the winners in an STV election were in winning positions in the first count.
But results, already mostly established in first round of STV vote count, are much different than FPTP or Block Voting results.
each voter has one vote in multi-member district -- that is backbone of STV.
okay, Cambridge uses really random, but that is not the only way to transfer without going to the extent of using Gregory method with its fanatical fractions.
Ireland and Malta uses Exact Method, transferring whole votes, not random at all as to first next usable preference. it has worked for 100 years.
And Gregory is not so pure --
Some Gregory methods use only "last parcel" to formulate transfers so sure it is not random but the trasfers it produces are not necessarily reflective of all the supporters who voted for the successful candidate. the transfer was done by using only certain of the winner's votes to formulate transfer. (it is complicated)
Cambridge likely is fine to use really random for these reasons ---
-sheer randomness is likely proportional. odds are anyway;
-votes are not examined so see how off it may be;
-seldom does a few votes here or there make a diff.
despite randon-ness, Cambridge results are mixed, balanced and roughly proportional.
mostly this is done in first round before any transfers are done.
look at Vanuatu's elections - no transfers at all (SNTV) and result is dependably more balanced than any Block voting or FPTP.
every district elects mixed rep. no artificial regionalism/geographical polarization.
That shows how little effect transfers have.
But STV is better - it polishes the fairness even more.
But that polish is not worth having a fetish over.
The Exact Method using whole votes is perfectly fine to do surplus transfers. IMO
----------------
further
the "Cast Vote Record" under any STV election would show that a high proportion of cast votes were used to elect the winners, no matter how the few surplus transfers were done.
in Cambridge (Mass.) city election 2023, 20,920 votes actually elected someone out of 23,339 valid votes.
just slightly more than one quota were not used effectively.
all nine elected in the end were in winning position in the first round, which can happen or come close under any STV system.
(see Bass division, Tasmania 2024 state election where seven elected, using Droop and Gregory method.
Wood with 1900 votes in first round jumped ahead of Armstrong with initial 2000. but otherwise all seven were most popular in first round and went on to win. Wood was able to aggregate many votes because three Liberals were excluded and had their votes transferred, while only two Lambie candidates were not elected.)
Wood elected at the end with 1000 more votes than last unsuccessful candaitre so not likely even a random transfer in the six surplus transfers that were done would have had any effect on who won.
2023 Cambridge
23,339 valid ballots.
Electing 9 candidates.
Quota is 2334 votes. (1/10th of valid votes , plus 1)
There were 173 invalid ballots.
20,920 votes used to elect the winners.
16,846 were cast in first round for a winner,
so never transferred except maybe as a surplus vote.
Cambridge election method statement says every nth votes used for surplus votes. so apparently they use "non-random random" (Chicago style, I think it was called)
=========================================
תגובות