Including abstention as a choice in vote choice models enables one to calcu
late the vote-stealing and turnout effects of third-party candidates. A mod
el of the: vote including abstention also produces parameter estimates and
marginal effects for some explanatory variables that differ from the result
s of a model that excludes abstention. We present two multinomial probit mo
dels of vote choice in the 1992 U.S. presidential election. One model inclu
des abstention as a choice; the other does not. The model that includes abs
tention reveals that Ross Perot increased turnout by nearly three percentag
e points in 1992, and his candidacy decreased Clinton's margin of victory o
ver Bush by seven percentage points. Under compulsory voting, Clinton's mar
gin of victory barely increases, and Perot's vote share remains nearly cons
tant.