Tullock (1981)
has recently restated a discrepancy between theory and observation that exists in collective decisions, and made a start toward explaining the discrepancy. The discrepancy, stated simply, is this: The theory of rational action in collective decisions implies that any outcome which requires vote exchanges to be achieved is unstable in that it can be overturned by a coalition all of whom find it their interest to do so. Yet, a high degree of stability is observed in legislation, despite the fact that much of this legislation required vote exchanges in order to be passed. Tullock accounts for this stability in part by pointing out that defections necessary to create the instability (e.g., a defector induced to become an'aristocrat' in an inegalitarian coalition) create an exposed position for the defector and the possible instability is likely to inhibit the defection in the first place.
Tullock looks at two extreme forms that a winning coalition could take, given a set of bills each of which has a special interest for one legislator: a logrolling form which encompasses all of nearly all the legislature, and a minimal winning coalition. Although the latter dominates the former in a game -theoretic sense, Tullock argues that the latter will be unstable because of the power of enticement held by the excluded minority. Logrolling which includes nearly all the legislature is, although formally dominated, stable because of the absence of this potential enticement. Thus for Tullock -if I understand correctly his argument -the potential instability of minimal winning coalitions leads any one of them to be succeeded, not by another minimal winning coalition, but by a non-minimal winning coalition that it formally dominates.
Shepsle and Weingast in a subsequent paper argue that the institutional structure of actual legislatures restricts the richness of logrolling and coalition possibilities that under pure majority rule without such institutions could overturn a proto-coalition or reverse an outcome (Shepsle and Weingast, 1981: 511). Thus they explain stability by stability-inducing institutions.