The probability of a voter being decisive (P-D), that is, of one vote affec
ting the outcome of an election, has generally been incorrectly calculated
for the last twenty or more years. The method normally used is due to Banzh
af (1968) and generalised by Beck (1974). It assumes that voters know in ad
vance how many people will vote for each candidate, which is clearly not th
e case. The correct formulation was given by Good and Mayer in 1975, but wa
s ignored and has subsequently been all but forgotten since then. A simple
explanation of these methods is given. Using the incorrect method, errors o
f magnitude of more than 10(100) in calculating P-D correctly can be made.
The appropriateness of using a decision-theoretic formulation instead of a
game-theoretic one is also discussed.