[EM] lying to pollsters (was comparative effectiveness ...)

Forest Simmons simmonfo at up.edu
Thu Feb 3 11:12:32 PST 2005


> From: Russ Paielli <6049awj02 at sneakemail.com>
> Subject: Re: [EM] Comparative Effectiveness of Approval and Condorcet
> 	in the	case of a three candidate cycle.

<snip>

>
> Yes, Approval does have some nice properties under the ideal conditions
> of DSV, but let me play "devil's advocate" again and bring up some
> "real-world" concerns.
>
> I've already brought up the issue of inaccurate polling data, and I
> think the effect of such uncertainty needs to be addressed before the
> effectiveness of Approval can really be evaluated. Someone somewhere has
> probably addressed this issue, but I am personally unaware of it.


You are right that disinformation is Approval's Achilles heel. That's why 
I would recommend ignoring the official polls completely and only rely on 
previous election results and/or partial results in the current election. 
If neither of these is available, treat the election as a zero information 
election.  As I have shown this still gives you a two thirds chance of 
making a favorable (for you) difference if the election is close, and 
(more importantly) when the next election comes around it will not be zero 
info.



> Beyond that, has anyone considered the optimal "strategy" in responding
> to a pre-election Approval poll? Since Approval strategy is so dependent
> on polls, this could be a significant issue. Obviously, people are free
> to lie to pollsters. Some may consider that unethical, but many will
> have no such qualms.
>
> So what is the optimal strategy in responding to an Approval poll? Do
> any or all voters have an incentive to lie about their cutoff point --
> or perhaps to even rearrange their preference order before drawing the
> line? And how would such strategy affect convergence if everyone adopted
> it? Will honest respondents be at a disadvantage?
>
> Obviously that question is a lot easier to ask than it is to answer, but
> I think some sort of answer in necessary before Approval can be fully
> evaluated.
>

I think that the average person would be apt to play up support for his 
favorite and down play support for compromise, and that most folks would 
take this into account when interpreting the polls.

I don't think the problem is with respondents lying so much as with 
private corporate pollsters having conflicts of interest (to some degree) 
and (to a much greater degree) corporate news media selectively spinning 
the results while reporting to the voting public.


Better to start with zero info strategy, and build up data, election by 
election.

That's where Condorcet has an advantage over Approval: no lag time.


Forest





More information about the Election-Methods mailing list