operation within the party, providing an indicator of its distribution of power. The most decentralized or "democratic" method of selection would involve the direct election of the leader by party members of even party "identifiers" or supporters. The most centralized of "oligarchic" method of selection would involve the leader naming his successor. The range of alternatives between these points on the continuum might involve a variety of "open" or "closed" procedures, including out-and-out skullduggery. In this variable, we are less interested in the proprieties of the process than in the numbers and credentials of people who participate in the selection.

Operational Definition. The higher the score on this scale, the more "centralized" is the party on leadership selection.

0	No national party leader can be identified.
1	The leader is selected by vote of party identifiers or supporters .
2	He is selected by vote of party members, a smaller group than above.
3	He is selected by a national convention or congress, the delegates to which represent regional or local party organizations.
4	He is selected by the parliamentary delegation of the party.
5	He is selected by a national convention or congress, the delegates to which are appointed or primarily determined by the national party organization.
6	He is selected by the national executive committee or party council subject to ratification by some lower levels of the party.
7	He is selected by the national executive committee or party council without further review of the decision; selections include change in leaders as a result of "power struggles" within the top leaders group.
8	He is selected by his predecessor; selections include the situation in which the party was created the incumbent leader and no apparent means transferring leadership have been established.

Coding Results. Coding parties on their procedures for "selecting the national leader" proved to be relatively manageable. As revealed in Tables 10.2a and 10.2b, 80 percent or more of the parties were rated for BV902, the data quality codes averaged above 7, and there was no significant correlation between BV902 and AC902. Moreover, the operationalization captured considerable variance among parties in their manner of selecting leaders. While about half the parties featured centralization in the choice of leaders with little opportunity for rank-and-file participation (codes 7 and 8), more than one-fifth selected leaders through national conventions with the delegates representing local organizations (code 3). Note that the proportion of parties that gave the current leader the power to select his successor (code 8) increased somewhat from the beginning to the end of the 1950s This was because the new parties that arose during the second half of our time period had not yet confronted the issue of leadership change. Perhaps these newer parties should be filtered out of any analysis involving BV902 as an indicator of centralization of power.

Our concern with the distribution of control over the selection of parliamentary candidates stems directly

go to page 111