In single-member constituency, simple plurality electoral systems, the disproportion between the popular vote polled by a party and the number of representatives it elects has stimulated much controversy. We review the mathematical and statistical models of two-party systems of this type which have been proposed by Kendall and Stuart, Theil, Lemieux, March, and the present authors. This leads to a detailed empirical study of series of elections in three systems: the provincial systems of Quebec and Ontario, and the Canadian federal system. In verifying a “sliding distribution” model more general than that of Kendall and Stuart, we construct the distribution of constituencies according their level of support for a given party, and hence provide an index of electoral regionalization. Less detailed seat-vote data, for longer series of elections in five systems, provide evidence for the persistence of electoral disproportions and the distinguishability of seat-vote characteristics between different systems. The extreme sensitivity of legislative proportions in Québec to small changes in popular vote proportions is shown to be a consequence not of peculiarities in the electoral map, but of the small standard deviation and large kurtosis (peakedness) of the distribution of constituencies.

If the importance of one-party dominance as a determinant of the rise of third parties were to be evaluated on the basis of the attention it has already received from political scientists, we would have to conclude that it must not be negligible. Obviously, however, this would not be a proper way of evaluating a theory. Fortunately, its importance rests on other grounds. Apart from my own quantitative and qualitative evidence, there is now Graham White's evidence, presented above in his article “One-Party Dominance and Third Parties,” which, as will be seen, I read to be in line with my own. And while André Blais, in “Third Parties in Canadian Provincial Politics,” above, concludes that he was “impressed with the number of negative results of the tests [he] used” (I am not!), he nevertheless showed an unusual reluctance to discard the theory and finally maintained that “one-party dominance is not irrelevant,” though it is “not a crucial factor.” We shall see later that his evidence is far from being that negative. Another critic of the test of the theory at the district level maintained, nevertheless, that the theory itself was very sound (“inattaquable”) and that its tests at the provincial level were convincing.

One of the most interesting points raised in Thomas A. Hockin's introductory essay to Apex of Power was the apparent contradiction he noted between S.J.R. Noel's use of Arend Lijphart's consociational model to describe the “Canadian political network” and the cabinet, and George Szablowski's interpretation of the same political ralationship through the application of F.G. Bailey's models of transactional and bureaucratic leadership. “The traditional transactional interpretation of prime ministerial-cabinet relations emphasizes each Minister's personal (mostly regional) ‘representative’ role in getting into the Cabinet and explaining his leverage within it … It may be in keeping with S.J.R. Noel's identification of the necessary attributes of the consociational political network in this volume. Yet George Szablowski in this volume questions whether this transactional dynamic is not threatened in Cabinet by a policy-making system which, with the aid of programme budgeting and clear leadership from the Prime Minister in the Cabinet Committee on Priorities and Planning, emphasizes rationality and the supra-regional functions of policy.”

From my perspective, two interesting questions arise from Stark's comment: first, a question concerning the usefulness or explanatory power of Lijphart's élite accommodation theory; and second, a question about the changing nature of infra-élite relations (specifically those occurring within the federal cabinet system) as distinct from inter-élite relations, primarily at the governmental level in Canada. Bailey's model becomes relevant in the discussion of the second question.