[EM] Condorcet cycles in Romanian 2009 election?

[Juho]

On Dec 10, 2009, at 2:36 AM, Warren Smith wrote:

> --It seems to me the Romanian O>B>G>O cycle was not "weak" in the
> sense it was strong enough to be stable day to day, and there was some
> nonrandom "rational"
> reason behind it.   However, it wasn't very strong in the sense the
> margins were about
> 3.5%, 0.66%, and 7.7% respectively.

In a stable and "opinion conservative" society (or when at least  
opinions concerning one of the topics are stable) already a 0.66%  
margin may be reasonably strong and stable. People may have some very  
stable opinions e.g. on religion, left vs. right, liberal vs.  
conservative and ethnic questions.

- - -

Here's btw also one concrete example of a possible stable cycle. Let's  
take first a traditional set-up on a one dimensional (left-right)  
political space. We have three parties, left (L, 47%), centre (C, 10%)  
and right (R, 43%). We may assume that C>L (since almost all R  
supporters will support C). We may assume that L>R (since about 50%t  
of the C supporters prefer L to R). Then we need some additional  
reason why R>C. In the basic linear scenario C would win R, so there  
must be some additional reason. Let that reason be that the candidate  
of C party has earlier given some strong negative statements about the  
leftist values of party L supporters. The candidate of party R has  
been more diplomatic. This is enough to make sufficient number of  
party L supporters dislike the C party candidate so much that they  
will rank her last. As a result R>C. Those old statements of the C  
party candidate do not change the other preferences. They may have  
actually made the C>L preference even stronger. C party supporters may  
still like L (as much as they like R) although their nominated  
candidate maybe doesn't always feel that way.

In an opinion space that focuses on multiple separate questions strong  
and stable loops may be even easier to arrange. There are three  
parties, right (R), left (L) and green (G). All are about equal in  
size. These parties are not on a linear spectrum but rather form a  
triangle with equal distances between all parties. The R party  
candidate happens to be slightly L oriented and not G oriented. Also  
candidates of the other parties have been nominated in a similar way,  
not from the very centre of each party but they all have similarly  
somewhat biased opinions. Together these biased opinions of the  
nominated candidates will form a cycle (with some not so small  
probability, assuming that the leading three parties are roughly equal  
in size).

In real life opinions are of course not as clear cut as in these  
examples. But if there is some bias among the voters/candidates in  
this kind of "cyclic direction" then the end result may well be a  
stable cycle.

Now when rethinking about the definition of a strong or stable cycle,  
maybe the characteristic feature is the stability. A cycle can be said  
to be stable if one can collect sufficient justification that explains  
why it is likely to be and stay stable (can be based e.g. on available  
polls (that cover a sufficiently long time span) or a theoretical  
credible model that explains why people feel this way).

Juho