You don't need to count them all in the same place. You could conduct each round of counting at the district level. Then, when the round finishes, you add the district counts together to see whether there is a candidate with a majority. If not, you eliminate the last placed candidate from all districts (i.e. the same candidate from each district), redistributing votes for him/her where possible, count the redistributed votes and add them to the last sum.
You have a pile for each candidate. The pile for the eliminated candidate is redistributed to the other existing piles in the following way. Each redistributed ballot paper is directed to the highest preference marked on it that is still in the race, if any. Only the redistributed votes need counting at this stage. You don't combine the redistributed votes with the existing piles until you have counted them.
I understand this. Which means that there needs to be these many actual physical recounting.
I believe an explanation is in order - in India we have up to 20 candidates in constituencies. We don't have a 2 party system. It's unfeasible if paper has to be counted 10-15 times.
Which is why I wanted to figure out if there was a one time counting system that lets us reassign votes without the need to physically recount again.
Well, papers with a first preference for any candidate that stays in the race until the last round of counting are only counted once, which might save a lot of time with many candidates if there is one very popular candidate (and even if there is a higher small number of popular candidates).
With large numbers of candidates, voters probably leave more unranked. When there are no more valid preferences on a ballot paper, there is no further need to count it. (I would prefer their votes to go to compromise candidates, instead of being exhausted, though.)
It depends on circumstances, but I expect votes for the candidates to be eliminated earliest would fairly quickly be redistributed to compromise candidates.
Consider an election between n candidates who are left/right-of-centre to varying extents. If my first preference is for a candidate at one extreme and my subsequent preferences are in order of similarity to that extreme and there are n-1 rounds of counting and my vote is counted every time, then my vote is probably quite unusual, because my favourite candidate was eliminated first and must therefore have had the fewest first preferences. However, the election itself is unusual, because the eventual winner is the kind of decisive fringe candidate that usually attracts few votes other than first preferences, but if such a candidate won on the strength of first preferences alone, there would be only one round of counting and no ballot papers would be counted more then once.
I don't know what the upper limit is on the number of counted preferences (as a function of the numbers of candidates and voters), but I would be interested in finding out.
I spent some more time on this - for a large number of candidates (which is extremely common in India), the IRV mechanism is impractical.
The Indian elections are the largest elections of any kind in the world. The number of eligible voters are 814 million and a total of 8k candidates contested for the 500 odd seats.
Thanks for the interesting links, but I don't think they support your conclusion that IRV is less practical to count. The Kemeny-Young method can't practically be counted without a computer. IRV can easily be counted by hand or computer.
I think IRV is used as widely as it is not only because of its simplicity, but also because of familiarity; it is a preferential form of the multi-round Exhaustive Ballot system and a special case of Single Transferable Vote (which in other cases is a system of proportional representation).
For your legislative elections, a system of proportional representation would be better than IRV, Kemeny-Young or any other single-winner system.
> For your legislative elections, a system of proportional representation would be better than IRV, Kemeny-Young or any other single-winner system.
that is a much harder battle. Changing the election system is easier than change the representation system.
So I agree that IRV is simpler - but I'm still looking for a tabular format that will allow count-once-but-execute-multiple-reassignment-round mechanism.
Sorry - I'm really interested in these things for the electoral system in India... which is far more complicated than the two party systems in the US.
