Over two centuries of theory and practical experience have taught us
that election and decision procedures do not behave as expected.
Instead, we now know that when different tallying methods are applied to
the same ballots, radically different outcomes can emerge, that most
procedures can select the candidate, the voters view as being inferior,
and that some commonly used methods have the disturbing anomaly that a
winning candidate can lose after receiving added support. A geometric
theory is developed to remove much of the mystery of three-candidate
voting procedures. In this manner, the spectrum of election outcomes
from all positional methods can be compared, new flaws with widely
accepted concepts (such as the "Condorcet winner") are identified, and
extensions to standard results (e.g. Black's single-peakedness) are
obtained. Many of these results are based on the "profile coordinates"
introduced here, which makes it possible to "see" the set of all
possible voters' preferences leading to specified election outcomes.
Thus, it now is possible to visually compare the likelihood of various
conclusions. Also, geometry is applied to apportionment methods to
uncover new explanations why such methods can create troubling problems.