As it happens, I've never paid attention to the details of how PR-STV works. So, in a sense, my mind is free of distraction on the point, and what I come up with *may* represent an intuitive approach of some value. If my intuition is sound, it may also match what has come to be seen as a more mature PR method. Or not. In any case, there may be some pedagogical value to be extracted from these considerations.
First of all, when I first heard of "Single Transferable Vote," I misunderstood it. I misunderstood it in a way that actually suggests what I'd call an Advanced Election Method. Such methods are ones that introduce serious reforms.... such as making direct democracy practical on a large scale. What I (mis)understood was that votes went to candidates at the voter's choice, and that any excess votes going to a candidate could be transfered *by the candidate.* Likewise, any votes short of a quota could be transferred or combined with other such vote groups, in order to create winners. This, of course, is a simple form of Asset Voting. My misunderstanding -- which seemed so attractive to me -- predates my contact with Warren Smith who formalized Asset. But as soon as I knew that STV had a ranked ballot, well, there goes that idea! (As far as being something that was actually being used anywhere.) Okay, so what then? Well, the obvious idea is that when a winner has been created, additional ballots being counted look at second place choices. This, however, has the obvious objection that winners can then depend upon counting sequence. And one reasonable solution, to a degree, is to count the ballots in random sequence, which, overall, would *usually* -- except when it is close -- choose fair winners. But this suffers from the problem of not being reproducible (it's possible to make it reproducible, but .... at the cost of additional complications). So what to do? Well, the next obvious idea is to fractionally distribute the votes, so that the collection of voters who voted for A in first place, all of them, if there are excess votes for A, have fractional votes remaining to be cast for their second place choices. The result is that each person has cast a sum of votes equal to one. So if a quota is 100 votes, and A gets 150 votes, each ballot for A becomes, now, 1/3 of a vote, 2/3 of each ballot having been used to elect A. So the second choices on those ballots get 1/3 of a vote each. And then this is applied recursively, when second choices are elected, etc. Seems fairly simple to me? Does it match actual usage? In this scenario, it might be tempting to round off the numbers. I see no reason at all to do this (at least not to round them to the nearest whole vote. Maybe to 1/1000 vote or some other fraction of a vote.) At this point it's only numbers.... and the skill involving in adding, subtracting, and dividing isn't all that great! The procedure also allows all the ballots to be counted first. So it's reproducible for audit. Essentially what one wants is a list of the ballot types, with the number of votes for each type. Easiest, I'd think, if the list is categorized by first place vote, then second, etc. Turning this list of votes into a list of winners could easily be done by hand, unless the number of candidates gets large. The ballots could also be counted sequentially, as needed. I dislike this, because I think every vote should be counted, even if supposedly "moot." If I went to the trouble to cast it, it shouldn't be tossed in the trash! If I was a candidate for office, and it turns out that many people voted for me, but not at a high enough preference for me to be elected, I'd hate not to know this! The result might actually be encouraging. Or not, depending on what is in those buried votes.... ---- election-methods mailing list - see http://electorama.com/em for list info