The (Un)predictability of Primaries with Many Candidates: Simulation Evidence

It is common to describe the dynamic processes that generateoutcomes in U.S. primaries as ``unstable'' or ``unpredictable''. In fact, the way we choose candidates may amount to alottery. This paper uses a simulation approach, assuming10,000 voters who vote according to a naive, deterministicproximity rule, but who choose party affiliationprobabilistically. The voters of each party then must choose between twosets of ten randomly chosen candidates, in ``closed'' primaries.Finally, the winners of the two nominations compete in thegeneral election, in which independent voters also participate.The key result of the simulations reported here isthe complete unpredictability of the outcomes of a sequence ofprimaries: the winner of the primary, or the party's nominee,varied as much as two standard deviations from the medianpartisan voter. The reason is that the median, or any othermeasure of the center of the distribution of voters, isof little value in predicting the outcome of multicandidateelections. These results suggest that who runs may havemore to do with who wins than any other consideration. If more than two parties or candidates are expected, then thevote-maximizing position is not close to your opponents, butwell away from them. (Tullock, 1967: 55).