The Laver and Shepsle (LS) model assumes that coalition government partners
can never co-operate to agree policy jointly, but must allocate ministeria
l portfolios between them, with each side's ministers acting as partisan 'b
arons' on their departments. Portfolio allocations do indeed make some init
ial difference as a stage through which all coalition negotiations must pas
s. but in the LS pure model their empirical significance for a coalition's
eventual policy stance is greatly exaggerated, ignoring the importance of a
range of governance mechanisms in fostering policy coalescence. Theoretica
lly, we should also expect two rational parties always to offer policy stan
ce concessions in bargaining, which move away from the LS 'lattice points',
in many circumstances maximising their joint welfare. Pressure from voters
, activists, party backers and the media will be for a coalition deal more
comprehensible than the lattice points for non-lattice audiences, located s
omewhere along the contract line from party A's optimum to party B's optimu
m.