Path: Janda: Political Parties, Home Page > Part 1: Table of Contents > Chapter 3

Chapter 3: Institutionalization (pp 19-28), this is p. 27
(you can navigate to other pages by clicking on page numbers below)
p. 19
p. 20
p. 21
p. 22
p. 23
p. 24
p. 25
p. 26
p. 28

computed from the average AC codes used in scoring the parties for the seats they held in each year of the period. For example, if we were sure about the proportions of seats the party held in half the years--assigning an AC of 9 to each year, but less confident about the proportions for the other half of the period-assigning only an AC of 7 to those years, the AC105 code for that party would average to 8. That average AC value for scoring each party on legislative representation became the AC105 code for legislative instability. In no instance did the average AC code for any party drop below a value of 4, indicating that relatively good legislative representation data existed to provide anchor points in calculating BV105.

Basic Variable 1.06: Electoral Instability

Political parties interact with the citizenry by contesting elections. The stability of these interactions should be reflected by the fluctuations in the percentages of votes received by a party over time. If a party never participated in an election during our time period--because of prevention or boycott--it cannot be credited with building electoral relations with the citizenry. Indeed, it is argued that the party's repeated absence in elections contested by other parties further disadvantages it relative to those which conducted campaigns. Thus, absence from two elections should reflect more instability than absence from one and so on, with the increment. in instability diminishing with each missed election. Of course, those parties which compete in some elections, but not all, should gain more credit for their participation than those absent from all the elections.

Operational Definition. Two different formulas are used to compute electoral instability, BV106. For parties which participated in at least two elections, BV106 is expressed by the coefficient of relative variation in percentage of votes (Yi) won in elections for the lower house of each of K elections during our time period as given by the formula

Electoral Instability =
Coefficient of Relative Variation =

Wherever possible, elections for the lower house of the legislature were used in applying this formula. In certain cases, however, we substituted votes cast for the parties' candidates in presidential elections. (The source of votes is identified in the country printouts.)

If this formula were applied to parties which did not contest any elections during our time period, the resulting instability score would be zero--for there would be no variation in the parties' "votes" not received. For these parties, therefore, an alternative formula is used to calculate BV106:

Electoral Instability =

where K again is equal to the total number of elections during our time period. Thus a party that missed the only election held would receive a score of 1; missing both elections when two were held would produce a score of 1.5; missing three out of three would produce a score of 1.83; and so forth.

If a ruling party consistently won large proportions of the vote in elections due to restricting competition by other parties, it would obtain a low electoral instability score according to our formula. In contrast to our policy for treating legislative instability, however, we do not credit such a party with a low electoral instability score. We regard a low electoral instability score in this situation to be markedly more artificial in character. While these electoral results are widely recognized as fiction, the percentage of seats that the party wins is clearly accepted as fact. Moreover, such a party is certain to be rewarded for its electoral policy by a low score on legislative instability (BV105) and to give it a low electoral instability score as well results in a form of doublecounting. Therefore, we omit computing BV106 for any party that obtains more than 90 percent of the vote averaged over all elections. This figure effectively filters out the parties as desired.

Coding Results. The scores computed for BV106 ranged from 0 to 2.08. They are grouped into categories for presentation in Table 3.8. As the "missing" data count shows, our success in coding parties on BV 106 was only slightly less than that enjoyed for BV105. While roughly 8 percent of the parties failed to be coded for both "electoral instability," and "legislative instability," these tended not to be the same parties. Parties that were banned throughout our time period and thus were not "eligible" for a legislative instability score were still coded on electoral instability if the country held elections, as described in the conceptual and operational definitions. But, when a party was assigned an electoral instability score for failing to participate in its country's elections, rigged or not, the resulting BV 106 score was always automatically given an AC score of 3--the lowest that can be assigned to any variable. Partly as a consequence, the mean AC code for BV 106 is considerably lower than that for BV 105. The lower confidence in coding electoral instability as opposed to legislative instability was also due to less attention paid in the literature to electoral success in comparison with legislative representation. There were thus two conse-

go to page 28