[EM] Majority? Expressivity? Strategy?

[Joe Weinstein]

MAJORITY?

Demorep sometimes deftly clarifies EM-issues. On the issue of 'majority', he 
has just written (3/29/01):

"Approval has the elementary defect of permitting a *real* first choice 
majority winner to lose (if *real* rankings were being used)."

Demorep gives a 100-voter example, namely Approval vote-counts together with 
their voters' 'sincere' underlying grades (on scale 0-100):
51 A (100) B (99)
3 B (100)
46 C (100)

Demorep's plaint is that, using Approval, B wins [with 54 votes] "even 
though A has a *real* first choice majority."

Precisely!  Only, rather than a defect, such behavior is a real plus for 
Approval or - even more - higher-resolution grading.  [Approval allows the 
voter to award each candidate, independently of all other candidates, either 
of the grades 0 (fail) or 1 (pass); a higher-res grading method allows more 
grade values.]  With any of these methods, each candidate's usual score - 
the candidate's grade average, over all voters - is a measure of something 
like that candidate's  'social utility' or overall acceptance; the winner is 
required only to have maximum attained score, rather than somehow represent 
a first-choice 'majority'.

A true 'majority' is to be generally expected only when there are two (or 
fewer) candidates.  A 'majority' requirement makes sense only if you believe 
that of right there ought always to be at most two parties.  How is the 
political system supposed to get a meaningful 'majority' when voters are 
split among three or more truly distinct parties that have little or no use 
for each other?  [For instance, each party may differ from all the others in 
one of three or more distinct but crucial policy dimensions, e.g. individual 
rights, fiscal policy, and fed-vs-state powers.]

The holy grail and battle cry of 'majority' are not only Demorep's but also 
IRV's.  For an N-candidate election IRV insists on producing an inevitable 
if meaningless 'majority' after N-2 imaginary runoffs; as Forest has noted, 
one might as well speak of the inevitable result after N-1 imaginary 
runoffs, i.e. 'unanimity'.

EXPRESSIVITY?

When one points out the above and other complications and crazinesses of 
IRV, as vs. Approval, one wonders why any unbiased newcomer to election 
methods might still prefer IRV. There is one quite simple evident reason: 
IRV at least lets you express three (or more) signficantly different (to 
you) layers or grades of preference.  With IRV, but not Approval, you can 
express that (for you) Nader>Gore>Bush - with each preference difference 
being (for you) significant.

That's why I continue to support use - in large-scale computerized elections 
- of one or another higher-res grading method (e.g., grading on a 0-100 
scale) rather than simply Approval, the best two-grade method.  The main 
voter frustration with the prevalent lone-mark method is that one cannot 
express fully how one evaluates all the candidates: one must simply declare 
one as better than all the rest (whether as a vote for the 'lesser evil' or 
as a 'sincere' vote for an alternative), and quit.  But Approval - and IRV, 
and yet other methods much discussed on this list - also do not allow the 
full expression of voter evaluation which is nowadays quite feasible.

Two minor arguments (from Mike O. and Bart I.), not earlier rebutted by me, 
have been made against higher-res grading.  One is that it is too difficult 
and costly to implement (a 'heroin addiction', I think Mike called it at one 
point).  The same argument could and should be made against all other 
non-two-grade methods (i.e. every usual method other than lone-mark and 
Approval).  All of these others require computerized input and tabulation - 
but ARE feasible (essentially at zero or insignificant one-time extra cost) 
with today's (or next year's) computers.  When computerized balloting is 
deemed unaffordable or too cumbersome - as in many small or collegial 
non-public elections - Approval makes perfect sense.  Another argument is 
that one can suitably 'simulate' a middle vote, between 0 and 100, by using 
random means to decide whether to vote 0 or to vote 100.  This argument is 
feeble: 0 and 100 may each be quite distinct from the vote that the voter 
actually wants to cast, e.g. 50. (This desire to vote 50 may legitimately 
arise from any of various possible states of voter knowledge or ignorance 
concerning the candidate.)

STRATEGY?

The most sophisticated argument given in these postings against higher-res 
grading [at least as v. Approval = two-level grading] is that of so-called 
'strategic collapse':  'strategically', a higher-res method allegedly calls 
simply for voting as one might in Approval, i.e. giving every candidate 
either full pass or full fail.

But 'strategy' is being thus used in a very narrow sense: one's vote helping 
determine the winner. In fact probability of - and consequent expected 
utility from - such an effect is usually next to vanishing.

TRULY strategical voting by any given voter will maximize that voter's 
expected utility.  Contrary to narrow uses in EM-list postings of the term 
'strategy', generally only a small component of this utility - let alone its 
entire value - will depend on the joint effect of who wins the election and 
of the tiny probabilities for the given voter's vote to influence this 
victory.  If the narrow use of 'strategy', and its underlying version of a 
voter's 'utility', were truly valid, most voters should rationally stay 
home.

The probability of casting a decisive vote may be extremely low; but every 
vote (with a properly responsive scoring and reporting system) helps send a 
message for future politics to consider.  Many voters, when they think about 
it (possibly after being reminded of the probabilities) will maximize their 
vote's utility (or 'waste' it least) by trying to maximize the vote's 
message-impact; for instance, by giving highest grades to candidates - 
howbeit 'dark-horse' - with worthy positions, and middling grades to other 
acceptable candidates. Other voters will perceive highest utility in being 
able to say to history that they voted as precisely as possible in accord 
with their feelings or conscience.

When Forest et al confessed that last November they voted their 'conscience' 
or 'sincerely' - as if the opposite of 'strategically' - I was reminded of 
how the cleverest pragmatic mid-20th century economists took pains to try to 
square their consciences and supposedly 'unscientific' advice with 
'scientific' orthodox economics - which in those days insisted that you just 
had to have bad times in order to get good times.

True strategy will often coincide with - and anyhow NEED NOT be the opposite 
of - expression of conscience or sincerity.  For each voter, it all depends 
on what values and prospects most animate and impact that voter's utility.

Some EM-list members urge methods which limit voter expression: in order, 
benignly, to help protect the voters from impulses to supposedly bad 
behavior, e.g. excursions from supposedly rational strategy.  Well, such 
protection is well served by staying with lone-mark.  I see little point in 
going to the trouble of advocating other methods which will still needlessly 
limit voter power to express preferences. (I also do not see the point of 
scoring methods - e.g. Condorcet - which do not systematically reflect the 
social-utility results of voter expression; but I leave more on that issue 
for another time.)

I see every reason to favor feasible and sufficiently simple but highly 
expressive balloting procedures, used with reasonably transparent scoring: 
e.g. high-res grading.  Given the election method, voters should be 
encouraged to use it expressively and as they choose, in whatever sense of 
'strategy', if any, that appeals to them.

Joe Weinstein
Long Beach CA USA

_________________________________________________________________
Get your FREE download of MSN Explorer at http://explorer.msn.com