[EM] Are cycles a problem in practice?

[Andrew Myers]

A lot of energy is expended on how to resolve cycles. Is this really a problem
in practice? One reason I ask is that having run a number of elections on my
election server, I find that cycles seem not to be an important issue. In fact,
not only is there no cycle involving the top candidate, often there are no
cycles at all, even with many candidates. For example, both times that I have
run the "democratic primary" election, the result has been a _total order_
on nine candidates. In the "hypothetical presidential election", with 24
candidates, there is a cycle involving candidates ranked 14-16 by beatpath, but
nowhere else. The "favorite movies" election, with 45 candidates, also manages to
totally order the first 13 movies before cycles are encountered. The worst case
is the "favorite ice cream" election, where cycles currently start at candidate #3,
but an unambiguous winner and runner-up exist.

You can check the results out for yourself at http://www5.cs.cornell.edu/andru/civs.

Is there literature suggesting that cycles are going to be a problem even
when voters are mostly sincere?

-- Andrew Myers