[EM] Hybrid ranking / approval ballot design

Rob Brown rob at karmatics.com
Thu Sep 2 13:25:55 PDT 2004

I was thinking that, while ranking all candidates is preferable (IMO) to 
approval voting (since there is almost no strategy involved in the former 
assuming a good tabulation method), making a ballot that can rank a large 
number of candidates is very difficult and/or expensive to implement, and 
might just require a touchscreen or something.

However, if all you need to provide is, say, 3 rankings, it is not so hard to 
make a simple ballot (even if there are many more than 3 candidates)...for 

  1st     2nd    3rd 
choice  choice  choice

 [  ]    [  ]    [  ]       George Bush

 [  ]    [  ]    [  ]       Wesley Clark

 [  ]    [  ]    [  ]       Howard Dean

 [  ]    [  ]    [  ]       John Edwards

 [  ]    [  ]    [  ]       John Kerry

 [  ]    [  ]    [  ]       John McCain

 [  ]    [  ]    [  ]       Ralph Nader

(note:  all you can do is put a mark in each box....no numbers)

The problem, though, is that if you aren't allowed to rank all of them (in 
this case, 1-7), you don't really fully express your preferences, and the 
effects of strategy and vote splitting and such creep back in.

So here is a new twist (well, maybe it's not new, but it's new to me! :) )

What if you allow people to check as many first choices as they want, as many 
seconds, and as many thirds (the rest being implicitly ranked as 4th).  This 
provides more information than approval voting (which is a special case of 
this method, but where there is only a first choice option), and more than if 
it only allowed you to rank your top 3.  Of course it provides less 
information than one that allowed a full ranking of all candidates -- but is 
the difference significant?

To extend this a but, you could even allow people to rank a candidate as both 
1st and 2nd (etc), which would effectively treat it as a 1.5 (it would place 
such a candidate above one that was ranked in 2nd, and below one that was 
ranked as 1st choice).

As for tabulating them, you simply use one of the condorcet methods, which of 
course know how to deal with equally ranked candidates.

One advantage to this sort of ballot design is that there would be no invalid 
ballots -- any combination of marks would be meaningful.

(BTW, I noticed that Dave Ketchum mentioned something about a 3 column format 
where duplicates are allowed, but I was confused by all the stuff about 
letters and numbers [mine only allows checks in each column], and it also 
seemed to imply that all candidates are listed 3 separate times.  Dave, if I 
have simply reworded your suggestion, sorry.)

Finally, you could even do the following:  if you only pick only one 
candidate, your ballot could be assumed to imply the rankings of all the other 
candidates, based on an average of all voters who rank your candidate first.  
I know, that is weird and radical, but it does make it so that those who do 
not want to think hard enough to get the most out of the system, can still 
simply vote for their favorite candidate while helping similar candidates.  
For instance, I could simply pick Nader as my top choice, which would imply 
that I like Kerry better than Bush (because more Nader voters prefer Kerry to 
Bush).  Of course, if I don't trust other Nader voters to think similarly to 
me (such as if Kerry is my first choice and Bush my second), I better go ahead 
and check some 2nd and 3rd choices and such.


Rob Brown

More information about the Election-Methods mailing list