[EM] Missed opportunity, historically

[Abd ul-Rahman Lomax]

In the U.S. there have been three major advanced single-winner voting 
systems tried.

Bucklin.
Top-two runoff.
Single transferable vote, recently called "Instant Runoff Voting" 
because it supposedly imitates TTR.

All of these have been viewed as reforms, with enthusiasm, by 
political scientists and others.

It has long been considered a defect in Plurality that there can be 
majority failure. People were familiar with direct democratic 
process, where a majority is *always* required to make a decision. So 
finding majorities was always considered a plus.

STV is a venerable voting system reform, but mostly proposed and used 
for multiwinner elections, where it's quite a decent method, though 
not ideal. The Ware method, i.e., STV single-winner, was long 
considered quite defective, I read scathing criticisms of it from the 
19th century.

But in a two-party system, STV does eliminate the spoiler effect. It 
breaks down badly with there are three major parties, and in 
nonpartisan elections, it appears to imitate Plurality. (This was a 
surprising discovery for me, from analyzing and comparing Top Two 
Runoff with IRV in recent elections. TTR reverses the winner from the 
preference order in the first round, about one-third of the time. IRV 
almost never does (in nonpartisan elections), and frequently fails to 
find a true majority. It does *not* emulate TTR.

Both Bucklin and IRV were sold as eliminating the need for real 
runoffs. That was hype. No method can do that, i.e., always find a 
majority with a single ballot. Sometimes the electorate simply is not 
ready to make that choice. TTR almost always finds a majority, and 
it's a real majority if write-ins are allowed on the runoff ballot. 
(That is, if there are significant write-ins, and a close election 
among the ballot candidates, there can be majority failure. But the 
choice to compromise and terminate the election at that point, with a 
plurality winner, seems to be solid and not challengeable. Only Asset 
Voting, which creates a possibility of finding a "represented 
majority" in further process, could fix this.

Asset Voting, while based on some fairly simple principles, has not 
been tried in any public elections, anywhere, so, while we should 
begin the coversation, the place for considering Asset Voting is in 
non-governmental organizations. Asset Voting is so vastly superior, 
in theory, to all other mere voting systems, that a major part of our 
overall voting system advocacy should be promoting Asset methods for 
use in NGOs. We will be preparing for the future, a very bright path.

As to public advocacy, only IRV has seemed to have serious traction, 
but this was a political accident, and was really aimed, originally, 
at paving the way for the use of STV in multiwinner elections, to 
create proportional representation. The original goal was PR, not STV 
or IRV. The choice to promote IRV was a tactical error, backfiring 
because of the very serious defects of IRV.

But TTR is widespread, especially in nonpartisan elections for local 
jurisdictions, plus some statewide contests.

I have not studied the history of TTR in public elections. I do know 
that, where Bucklin, long-term in party primaries, seemed to not find 
majorities, with common bullet voting, it was eventually replaced 
with top two runoff. We should not go backwards.

An opportunity was missed when Bucklin was replaced with TTR. 
Instead, it would have been superior to leave Bucklin in place, and 
to hold a runoff when a majority was not found in the first round. 
Bucklin would still be quite superior to the vote-for-one that became 
the rule in TTR. Once there was a cost to bullet voting, i.e., 
majority failure and therefore a runoff, I strongly suspect that 
additional approval rate would have increased.

One of the goals of a party primary is to assess the relative 
strength of candidates. Vote-for-one only tests first preferences. 
The data from that first ballot would be better if it's a ranked 
ballot (or a ratings ballot as in Range), and would have been 
improved simply by Counting All the Votes. I.e, Approval as the first round.

Range methods, I have seen, are excellent polling methods, I saw, in 
2008, Range 2 polls (votes of -, 0, +, with individual abstentions 
being 0) that were very accurate, showing what vote-for-one polls 
could not show, strength of opposition as well as strength of 
support. Those polls showed Obama as being clearly ahead with the 
Democrats, with Clinton close behind. Clinton had many more positive 
votes, but many more negative ones as well. In the Republican race, 
Ron Paul, after debates, was a clear winner. In vote-for-one polls, 
he did very poorly, probably because he was not perceived as a viable 
candidates, and so people didn't want to waste their vote!

Bucklin should have been left, with a runoff added. Bucklin, if it 
fails to avoid a runoff, has still collected far more data, and 
candidates will know their real support much better.

In public elections, Bucklin seemed to almost always find true 
majority support. It does this much more efficiently than IRV, 
because of IRV's elimination of votes. IRV is vulnerable to the very 
serious center squeeze effect (which can very much damage party 
primaries, resulting in the choice of an extreme candidate who cannot 
possibly win the general election; in general elections, the center 
squeeze effect can easily result in the election of an extreme 
candidate, whereas a centrist candidate would win in a direct 
contest, even by a large margin.

Top two runoff is, of course, also vulnerable to center squeeze, it's 
famous for it. At least, though, a real runoff will result in the 
best choice from the top two, who will generally be the less extreme 
of the two choices. When write-ins are allowed, it's possible to fix 
the issue, but then a new spoiler effect can be created.

The fix would have been to keep and use Bucklin in both the primary 
and the runoff. Bucklin is not vulnerable to the spoiler effect. The 
runoff might well have only two candidates on the ballot, but voters 
can freely prefer a write-in. If there were only a single round, this 
could cause majority failure (part of the spoiler effect). But, then, 
with an additional approval, they could play it safe and vote for one 
of the ballot candidates.

We obtain a very powerful election reform by combining the most 
widely used election reforms. Bucklin was tried much more widely than 
IRV has recently been tried. TTR is quite common.

So, our strategy is twofold:

1. Promote Count All the Votes, i.e., Approval, as a very simple fix, 
cost-free, that can greatly reduced the spoiler effect, and that is 
clearly only a matter of giving voters a little more freedom of expression.
2. Preserve Top Two Runoff from attack by the Instant Runoff Voting 
people. It is superior in terms of election quality, and the cost 
savings, a major selling point for IRV, are speculative. It should be 
known that IRV, in nonpartisan elections, is an expensive simulation 
of Plurality. It absolutely not a reliable way of finding majorities. 
We should make known the evidence for this, it's not just an opinion. 
IRV has been deceptively sold.
3. Top Two Runoff with Count All the Votes becomes less vulnerable to 
Center Squeeze. It's a minor fix, and how much it will help is 
speculative, because it has not been tried. It's certainly not 
perfect, but a clear improvement, and more equitable.
4. Top Two Runoff with Bucklin ballots is a serious improvement over 
TTR with Count All the Votes, because it allows multiple approvals 
while still allowing ranking. This will return to use the benefits 
that we could have had, all along, if our system had synthesized the 
two reforms, the better part of a century ago, instead of considering 
them as competing.

Bucklin is Instant Runoff Approval. It's far more likely to find a 
majority in a single ballot, and then, when it doesn't, a Bucklin 
runoff is practically a dead certainty to do it. Without actually 
tying the voter's hands.

Bucklin ballots are cheap to count. If no additional approvals are 
added, it costs the same to count as Plurality. But we should simply 
Count All the Votes, the information is well worth the counting cost. 
Lower approvals should be counted even if the method has completed, 
has found a majority, say, in the first round.