[ska]

I don't see why paper makes different systems more difficult. Paper is more about how votes are registered and counted, voting systems are more about how they are aggregated and decided.

Re: your comment on social engineering, etc. There has been a ton of effort in arenas in the US electoral system(s), but the goals have not been security or access - sometimes quite the opposite.

[Mbioguy]

I was part of a now-tabled push for RCV in a red state a few years back. While it offered a solution to several issues and had some multiparty support, the big problem arose when the state legislators realized that the current machines used to aid in tabulating and reporting results couldn't do RCV tabulation. This meant either new machines would need to be bought or it would all have to be done by hand. The clerks ran the numbers and the number of volunteers required or increased cost was not viable. The decision was made to table considering RCV use statewide until the machines were scheduled to be replaced in half a decade.

The problem is that an individual ballot in RCV if counted manually, needs to be counted several times if the vote ends up switching to the 2nd, 3rd, or fourth choice etc. As the number of total votes increases from the thousands to hundreds of thousands (or even millions), this can cause a pseudo-exponential growth in the amount of counting labor performed. Large municipalities that have trialed RCV have forged ahead with manual counts and succeeded despite the increased effort, but it has sometimes taken days to report results. That's only after massive amounts of labor from volunteers.

Going statewide with manual-only counts of paper ballots (no optical scanning or other machine-assisted counting) makes RCV or other methods of determining winner much more difficult because of the labor or cost involved.

It should be clear that I quite like RCV, but this is a factor to consider.

[ska]

That's fair - technology limits will vary place to place. I would think that something like RCV would be best done in a single pass though, so you have all the information up front. But of course I've never tried to implement it! My initial though was it was approx 5x actual counts, but that's far less than 5x labor because you only do the logistical parts once. So I was thinking you could do it with about 2x the labor total, which sounds achievable.

Personally I'm a fan of just committing to doing it by hand and accepting the cost in exchange for robustness against many types of manipulation. The problem is your argument works the other way also, the easier/cheaper it is to tally the votes due to reduced labor and # of people involved, the higher the risk of significant interference, imo.