ational definition seem to have erred on the other side. The criteria for identifying purges are so restrictive that very few parties earn a code above 0 (see Tables 11.8a and 11.8b). According to our restrictive definition of a party purge, only 10-15 percent of the parties engaged in wholesale housecleaning of either members or leaders. Even if one is satisfied with our limiting the concept to mass expulsions and thus accepts the validity of our finding, BV1006 suffers as a variable in the comparative analysis of political parties, because it uncovers little variation in party practice. Although we were able to code more than 95 percent of all our parties on BV1006, although our judgments were made with relatively high confidence, and although there was no significant relationship between BV1006 and AC1006--the variable seems to be of doubtful use in comparative study. Whether there is any utility in distinguishing the small number of parties that engaged in purges from the many that did not, of course, constitutes a matter for research in the context of party theory.

return to Table of Contents