¹ Lowell, A. Lawrence, Governments and Parties in Continental Europe, Vol. 1 (Cambridge: Harvard University Press, 1896), pp. 73–74 Google Scholar.

² Ibid., pp. 70 and 73.

³ Parliamentary governments in countries whose elections were not fully competitive throughout this period are excluded. The data for Chile and Uruguay are not as comparably reliable as those used here. Switzerland and the United States are, of course, excluded by definition, since their executive is not dependent on the legislature. Our data are from Keesing's Contemporary Archives.

⁴ Blondel, Jean, “Party Systems and Patterns of Government in Western Democracies,” Canadian Journal of Political Science, 1 (1968), 180–203 CrossRefGoogle Scholar. This appears to be the only empirical comparative study to date of the relation between party systems and governmental stability; but no measures of the fragmentation of the pary system were used, so that correlational analyses were not possible, and only averages (over a twenty-year post-war period) were compared. In fact, if cardinal measures of fragmentation were used, the correlation of their over-time averages with the average duration of cabinets is found to be quite high, but we cannot make the desired inferences from this about the stability of single governments.

⁵ Macridis, Roy C., “Cabinet Instability in the Fourth Republic,” Journal of Politics, 14 (1959), 643–658 CrossRefGoogle Scholar.

⁶ Siegfried, André, “Stable Instability in France,” Foreign Affairs, 34 (1956), 394–404 CrossRefGoogle Scholar. See also Luethy, Herbert, France Against Herself, Trans. Mossbacher, E. (New York: Meridian Books, Inc., 1957)Google Scholar.

⁷ Lijphart, Arendt, The Politics of Accommodation (Berkeley and Los Angeles: University of California Press, 1968), p. 72 Google Scholar.

⁸ Only those parties holding at least two percent of the total seats in the parliament were included in this variable, and in all other party-system variables used below, except those which measure some aspect of the government, in which case parties with less than two percent of the seats which entered into the government were taken into account. Where figures for “others” and “independents” were reported with no further information, these were also excluded; very rarely did the size of such categories exceed the two per-cent level. We are aware of the dangers, in some types of analyses, of using a rigid cut-off point for minor parties. Sartori cites the example of the Italian Republican Party, which has had an average return since the War of less than two percent, yet has made up the majority in several government coalitions. See Sartori, Giovanni, “The Typology of Party Systems,” in Allardt, Erik and Rokkan, Stein (eds.), Mass Politics: Studies in Political Sociology (New York: The Free Press, 1970), p. 325 Google Scholar and footnote 5. However, the inclusion of these parties would not have changed the decisions made on any of our hypotheses.

⁹ This correlation is significant at the .01 level. All simple correlations in the paper are significant at least at the .01 level, unless stated otherwise. It should be remembered that the number of observations in every case is 196.

¹⁰ Lijphart, Arendt, “Typologies of Democratic System,” Comparative Political Studies, 1 (1968), 3–44 CrossRefGoogle Scholar. We would agree with Sartori (op. cit., p. 347 and footnote 59) that it is an example of the dangers of “predicting regularities on the basis of exceptions.” But in any case, we submit, the establishment of precise empirical relationships of fairly general validity is not materially helped by the use of mere typologies (of party systems), such as those proposed by Lijphart or Sartori himself.

¹¹ Rae, Douglas, The Political Consequences of Electoral Laws (New Haven: Yale University Press, 1967), pp. 53–58 Google Scholar. See also Rae, Douglas and Taylor, Michael, An Analysis of Political Cleavages (New Haven: Yale University Press, 1970)Google Scholar, chapter 2.

¹² Cf. Kirchheimer, Otto, “Majorities and Minorities in Western European Governments,” Western Political Quarterly, 12 (1959), 492–510 CrossRefGoogle Scholar.

¹³ Duverger, Maurice, Political Parties, Trans. R., and North, B. (London: Methuen, 1954), p. 415 Google Scholar.

¹⁴ “The opposition” is defined simply as all the parties not in the government.

¹⁵ Duverger, loc. cit.

¹⁶ See, for example, Stokes, Donald E., “Spatial Models of Party Competition,” this Review, 57 (1963), 368–377 Google Scholar; and Converse, Phillip E., “The Problem of Party Distance in Models of Voting Change,” in Jennings, Kent M. and Zeigler, Harmon L. (eds.), The Electoral Process (Englewood Cliffs, N.J.: Prentice-Hall, Inc., 1966), pp. 175–207 Google Scholar.

Later, we shall make an assumption about the distances between the parties along this single left-right dimension. This part of the analysis could (quite easily, in principle) be extended to incorporate multi-dimensional ideological spaces, with no restriction that the number of dimensions be the same in each country. For example, Leiserson, in testing coalition theories against data on cabinet coalitions in three European countries, used a one-dimensional space in Sweden, two dimensions in Italy, and three in France. See Leiserson, Michael, Coalitions in Politics (unpublished Ph.D. Dissertation, Yale University, New Haven, 1966)Google Scholar, chapter 7. But the identification of these ideological dimensions for every one of our 19 countries over the 24 years period is a task we shall not attempt here.

¹⁷ D is adapted from a measure proposed by Leik, Robert M., “A Measure of Ordinal Consensus,” Pacific Journal of Sociology, 9 (1968), 85–90 CrossRefGoogle Scholar.

¹⁸ Cf. Sartori, op. cit.

¹⁹ Duverger, op. cit., p. 419.

²⁰ Cf. Dahl, Robert A., “Patterns of Opposition,” in Dahl, (ed.), Political Oppositions in Western Democracies (New Haven: Yale University Press, 1966), pp. 344–347 Google Scholar; and Sartori, op. cit.

²¹ Axelrod has proposed a similar hypothesis, although his is derived from a theory of coalition formation which is not of the “rational choice” variety. From the assumption that the less “conflict of interest” there is in a coalition, the more likely it is to form and the longer it will last, together with his definition of “conflict of interest,” he derives the proposition that “minimal connected winning” (MCW) coalitions will form and will last longer than those which are not MCW. A coalition is “connected” if it consists of ideologically-adjacent members. Both parts of this proposition are successful when tested against Italian cabinet coalitions from 1953 to 1969, but no other data are examined. See Axelrod, Robert, Conflict of Interest (Chicago: Markham Publishing Co., 1970)Google Scholar, chapter 8. If we assume that the parties can be distributed along a single ideological dimension, and if we assume a cardinal measure of the distance between the parties on this dimension, then Axelrod shows (op. cit., chapter 7) that his measure of “conflict of interest” is identical with the variance of the distribution. Conflict of interest inside the government coalition is therefore the same as the variable V_G which we used in our analysis (although, unlike Axelrod, we assumed that the parties are spaced at equal intervals along the ideological dimension). Thus, Axelrod's hypothesis, that coalitions which are MCW (and therefore have minimum conflict of interest) last longer than those which are not, is similar to our hypothesis that there is a correlation between V_G and the duration of cabinets.