Democrats & Republicans should adopt Range Voting in their Iowa primaries
Give me a place to stand, and I shall move the Earth. —Archimedes.
We want to try to unify all voting reform forces behind a push to get single-digit range voting in at least one main party's Iowa 08 caucus. We summarize ten reasons for that:
In fact, essentially every player involved in Iowa 08 is motivated to want range voting. Let us check the list:
I attempted to resuscitate my old sim software to run it on an as-Iowa-08-like scenario as possible. Specifically, I did 9 candidates, 101 voters, and 9999 runs (i.e. all results are averaged over 9999 randomized elections) where the utilities are generated based on I "issues" each voter feels randomly about, plus a random perturbation.
| Voting system | Regret (0 issues) | Regret (3 issues) |
| strategic range = approval | 0.84 | 0.39 |
| strategic plurality = strategic IRV | 2.99 | 1.01 |
| honest range | 0.04 | 0.05 |
| honest approval | 0.59 | 0.28 |
| honest plurality | 1.94 | 0.76 |
| honest IRV | 0.94 | 0.45 |
| random winner | 4.36 | 1.48 |
(Errors in regret values are of order +-0.06 or less in first column, 0.03 in second.)
so as you can see in the Iowa-08-like scenario, it really really is a big win to use range voting. It wins a factor of 6 to 15 in terms of reduced regret versus approval, if voters are "honest." Based on my previous real world polling experience I think a rough estimate is 50-50 honest-strategic voter mix. (In reality a lot fewer than 50% are fully strategic but a lot of the nonstrategic ones appear to be partially strategic so I think 50-50 mixture is a decent approximate model.) In that case I estimate the regret with range will be about half the regret with approval (i.e. 50-50 mixture of "same" and "way less").
(What the numbers mean: "Bayesian regret" is explained "for dummies" here.)
The source of this really really big win is: approval tends to look less good if there are many candidates (here 9) and if a reasonable fraction of the voters are honest. (Indeed, with honest voters, approval is worse than Borda with 5 or more candidates.) Range voting allows taking advantage of honesty while not suffering from strategy. In Iowa we expect voters to be more honest than in most elections since they simply do not know who the "frontrunners" are, preventing them from being as strategic as most voters in most elections can be.
You can try to translate this all yourself into estimated probabilities that approval or range will select a different and clearly better candidate than plurality in Iowa 08, thus providing a huge boost for voting reform and a huge boost for whatever party was smart enough to use range voting in Iowa 08. Or, forget trying to do that yourself – simply consult our simulations that directly estimate the probability that the Range and Approval winners will differ (we find about 31% chance for an Iowa-like situation; or see our more general tables), and the probabilities that the Range winner will differ from the Plurality winner (we find over 80% chance of difference, in a random-election model).
Obviously the chances of that huge boost, are far better if we go with range voting, not approval. It is insane to sacrifice the opportunity for such a huge win for our team, and for the USA and world, (by "sacrifice" I mean "reduce chances by a factor of about 2") purely because of some kind of misplaced psychological confusion about approval and distractions. That stuff does not add up to a hill of beans versus reducing the expected regret of the entire USA by this much.