The electoral law and voter preferences are modeled as a partition-function
form game that represents payoffs of electoral coalitions in a multi-party
system. Three empirical hypotheses are formulated under the assumption tha
t seat-maximizing parties modify coalitional structures to extract gains fr
om coalescing. The hypotheses say that party systems under stable voter pre
ferences and electoral law tend to stability with respect to splits and coa
lescing. The partition function is estimated for various coalitional struct
ures for the Polish party system with simulation based on district-level el
ection results and survey data. The estimates show that the Polish party sy
stem is split stable. While the opportunities for gains from coalitions on
the Right were exhausted between the years of 1993 and 1997, some opportuni
ties for such gains emerged on the Left.