[Election-Methods] Reply to "parable" attacking me and/or range voting

Warren Smith wds at math.temple.edu
Wed Dec 26 16:06:24 PST 2007

Rob Brown posted a long screed purporting to be a "parable"
and somewhat subtly (or not) attacking me and/or range voting.
Unfortunately, it appears to have been based on misconceptions.
I do not want to get into a long debate, but I will take this opportunity to try to
correct the comon misconception(s) Rob's screed was grounded on.

Basically, Rob set up this scenario of people and apartments, and some fool
named "Warren" was trying to convince people it'd be far more "utilitarian"
if all the locks on all the apartments were the same.   

The analogy he hs in mind to range voting is, I presume, that either
(a) with range voting, anybody can exaggerate strategically in their vote, rating, say, 
Bush=99, Gore=0 and thus having far more power than some voter who rates Bush=50, Gore=51.
(b) With range voting, some fraudster can alter people's votes to say totally 
different things, thus getting a totally dishonest election result.

I do not know which one Rob Brown had in mind, but either way, he is not appreciating
the true situation.  Let me attempt to educate.

---Response to a---

First of all, even if EVERYBODY strategically exaggerates to the MAX, then range voting
just becomes "approval voting" which is still a good voting system.  Not a failure.

In fact, approval voting works better than instant runoff and Condorcet systems with 100%
strategic voters do, according to my computer simulations.  One reason:
with 100% maximally-exaggerating strategic voters, IRV and Condorcet and plain-plurality
ALWAYS elect the same person - who ALWAYS will be one of the two major-party candidates
(in the absence of a perfect tie).
This leads to permanent 2-party domination, and third-party candidates never being elected.
Which is a substantial part of the reason that in all the over-500 IRV seats in Australia,
only a single one was held by a 3rd-party member
(zero in the federal house last 2 elections despite every IRV
seat being contested by at least one third-party, usually several).

[E.g see    
to see that Australian politics is massively 2-party dominated just like the USA, and see
for clear evidence Australian voters vote massively strategic-exaggerratingly in style.
About how Ireland and Fiji also got 2-party dominated in IRV seats, see
http://www.rangevoting.org/FijiPol.html ]

Meanwhile, range voting, even with 100% maximally-exaggerating strategic voters,
can easily elect a 3rd party candidate by a clear gap.   Imaginary example:
 RANGE VOTING: 51%: Gore=99, Nader=99, Bush=0.  49%: Bush=99, Nader=99, Gore=0.
Result: Nader wins easily (and also would even if the voters had wimpily decided
not to exaggerate and only given him 60!)
 IRV OR CONDORCET:  51%: Gore>Nader>Bush,  49%: Bush>Nader>Gore.
Result: Gore wins.  Nader can NEVER win in this kind of situation.

So I guess I'd have to ask Rob Brown, assuming he prefers IRV, Condorcet, or
Plurality over range voting - why do you think only 2 parties should ever have a chance?
Might democracy work better if voters had *more* than the minimum possible amount of choice?
(For example, last presidential election voters could choose between the pro-war, pro-NAFTA, 
pro-WTO, pro-PATRIOT-act, pro-huge-farm-subsidies candidate, or the other one.  
Might the USA have been better off if the voters at least had a *choice* about those issues?)

Second, if NOBODY exaggerates, then range voting works better than all the other common
voting systems, in my computer simulations, and also the honeybees like it.  

Third, you might worry that perhaps if SOME range voters
exaggerate while some do NOT, then we'll get problems.
That sounds plausible at first, but when you actually do the experiment - see these
computer sims:  http://www.rangevoting.org/StratHonMix.html
you find out that actually, range voting still outperforms the common rival systems
even under circumstances designed to make it maximally suffer from this.

---Response to b---

True, fraudsters can alter range votes.
But they also can alter votes in any other voting system.
So this is not a deficiency of range voting per se.

Range and approval are actually somewhat more-immune to fraud than many other systems, 
for these reasons:

1. Fewer spoiled ballots:  see 
for the fact the best system (in terms of "spoiled ballot" rates from actual human voters)
is Aproval, 2nd best is Range, 3rd best is PLurality, and worst is IRV.
IRV actually has probably the greatest spoilage rates in the world.

2. Range and Approval work with Rivest's "ThreeBallot" revolutionary antifraud voting protocal.
Plurality and IRV do not.

At first I thought this was a big edge for range voting, 
but then it diminished when Rivest & Smith came up with 
the "VAV" and "Twin" protocols which allow  plurality voting and (if there
are only a few candidates) also IRV & Condorcet voting.  See 
for an introduction to Rivest-Smith secure voting protocols.

3. No such thing as an "overvote" with range & approval (but is with IRV and plurality).

4. Range and Approval share plurality's virtue of being "countable in precincts."
IRV cannot be counted in precincts. That's a big security hit.  For what I mean by that see

---More response---

Instant Runoff Voting (IRV)
also suffers a lot of nasty pathologies like "non-monotonicity", "no-show paradoxes", 
"fails to elect a Beats-All Winner" and "favorite betrayal"/"spoiler" effect.
Many foolish IRV-lovers are unaware of these problems, and wrongly think IRV abolished them
(or, if confronted with counterexamples, wrongly think they are extremely rare).
Also, IRV has historically led to 2-party domination in every country inwhich it
has seen extended and heavy use.
IRV pathology survey:

IRV non-monotonicity frequency calculation:

IRV failures in real life:
(a) Irish 1990 presidential election: the only IRV election in all of Irish history in which
the plain-plurality winnerdid not win, i.e. in which IRV had an actual effect.  And it
was non-monotonic - if Lenihan voters had chosen to vote aganst Lenhan, then 
Lenihan would have won:

(b) Lousiana 1991 governor (2-round runoff) election: fails badly, refusing to
elect beats-all winner, non-monotonic, etc.

(c) Australia latest round of IRV elections: there were many pathologies, documented here

The "nursery effect" which is an important reason to believe that range voting will
NOT lead to 2-party domination:

The fact (pointed out by Duverger)
that the (French, non-instant) runoff system generaly does NOT lead to 2-party domination:

The fact that Range & Approval voting work right now with every voting machine in the USA
with no modification needed:

In contrast, as IRV-advocate Rob Richie recently admitted in print, no voting machine
in the USA will handle IRV unmodified and out of the box.
Also, various IRV municipalities are now in hot water because of, e.g, California
decertifying computerized voting machines.   

Warren D. Smith
http://RangeVoting.org  <-- add your endorsement (by clicking "endorse" as 1st step)

More information about the Election-Methods mailing list