Previous studies of the factors that influence the ability of parties to join governments have estimated binary choice models using the parties as the unit of analysis, which inappropriately treats each party in a government formation opportunity as an independent observation (a problem that clustered standard errors do not solve) and does not allow researchers to control for important coalition-level effects. This article demonstrates that a preferred methodological approach is to first estimate a standard multinomial choice model (conditional logit or mixed logit) of coalition formation, using government formation opportunities as the unit of analysis and potential governments as the choice alternatives. The probabilities of parties joining governments can then be recovered by simply summing the probabilities for the potential governments that contain each party. An empirical example shows how the substantive conclusions about a party's likelihood of entering office can change depending on the methodological approach taken.