depend on the number of parties (at least under some electoral rules).8 Next we
discuss the possible outcomes for rGP and RP , while E(NsPG) is discussed in the
next subsection.
It is easy to show that the parties in government never find it optimal to confer
any rents on the parties in opposition. Rather than deriving this result in every
case studied below, we simply set rOP =0in the following. The first term in (2.4)
is thus relevant only for a party in government, i.e., for G = C, S.
The expected value of seats in the next legislature, RP , is taken as given when
policy is set in the current period, since it refers to future events beyond the
control of the current government. But as we shall see, the value of seats matters
not only for expected payoffs but also for current policy, in particular for the
optimal size of current rents, rGP . To pin it down, we let RP be proportional to
total equilibrium rents captured by the government, and we implicitly assume that
the party structure remains the same after the election. The expected value of
seats differs for a small party (P = J) and a large party merging two political
groups (P = IJ), because a small party can only be in a coalition government,
while a large party can only be in a single-party government. Specifically, let rGP
denote equilibrium rents captured by any individual party, conditional on being
in a government of type G = C, S. Then we assume
P ½ 2γrCP if P is small (P = J)
R = t YrSP if P is large (P = Ы) ,
(2.5)
where γ is a positive parameter measuring the payoff from future expected rents.
Thus, by (2.4) and (2.5), the parties want to win seats in the next legislature,
because they are associated with a bigger expected share of future government
rents. Although this formulation is quite plausible, we do not attempt to derive
it from more primitive assumptions about bargaining over government formation
in the next legislature. In this respect, we treat government formation as a black
box in the present as well as future periods.9
8 Note that we use different operators to denote expectations at the policy formation stage,
E(∙), and the party-formation stage, E(∙), since the nature of the uncertainty is different.
9 Preserving some symmetry, the results derived in the next section, including the equilibrium
party structure, hold for arbitrary seat shares in the initial legislature (and not just for equal ini-
tial shares among all parties). In this sense, the static equilibrium of this game is self-replicating,
and the implicit assumption that the equilibrium party structure remains unchanged after the
election is not too restrictive. Extending the model to an intertemporal set up, however, would
raise deeper questions about the payoffs and the rules for government formation. We leave this
extension for future research.