Computational social choice literature has successfully studied the complexity of manipulation in various voting systems. However, the existing models of coalitional manipulation view the manipulating coalition as an exogenous input, ignoring the question of the coalition formation process. While such analysis is useful as a first approximation, a richer framework is required to model voting manipulation in the real world more accurately, and, in particular, to explain how a manipulating coalition arises and chooses its action. In this paper, we apply tools from cooperative game theory to develop a model that considers the coalition formation process and determines which coalitions are likely to form and what actions they are likely to take. We explore the computational complexity of several standard coalitional game theory solution concepts in our setting, and study the relationship between our model and the classic coalitional manipulation problem as well as the now-standard bribery model.