where p_i is the share of the vote won by the ith party. In a perfect two-party system, with the votes split 50-50, ENPV equals 2.0; in a four-party system (25-25-25-25), it equals 4.0. But if some of the parties are larger than others, the effective number of parties is usually some fraction, usually a bit higher than the intuitively expected number of parties. The standard index of volatility (V) is the sum of all the changes in vote shares experienced by all parties from one election to the next, halved to eliminate duplication:

It ranges from 0 to 100 and can be interpreted as the percentage of the vote that shifts among parties, in the aggregate, between elections (Pedersen, 1979).

2. For the classifications themselves, tallies of the vote shares of each bloc, and a complete description of the methodology used in the classification of parties, see Coppedge (1997a).

3. I am deeply indebted to the country specialists who, without compensation, took the time and effort to comment on my draft classification. They are, for Argentina: Marcelo Leiras, James McGuire, Guillermo O'Donnell, Scott Mainwaring, and Edward Gibson; Brazil: Barry Ames, David Fleischer, Scott Mainwaring, and Timothy Power; Chile: Aníbal Pérez Liñán, Iván Jaksic, and Manuel Antonio Garretón; Colombia: Pablo Abitbol, Ronald Archer, David Bushnell, Robert Dix, Jonathan Hartlyn, Gary Hoskin, Francisco Leal Buitrago, and Steven L. Taylor; Costa Rica: Fabrice Edouard Lehoucq, Mitchell Seligson and Cynthia Chalker, Deborah Yashar, Manuel Rojas Bolaños, John Booth, and Jorge Vargas; Ecuador: J. Samuel Fitch and Andrés Mejía Acosta, who also supplied copies of published classifications by Fernando Bustamante, Luis Verdesoto Custode, and E. Durán; Mexico: John Bailey, Roderic Ai Camp, Robert Dix, Xochitl Lara Becerra, Soledad Loaeza, Alonso Lujambio, Kevin Middlebrook, Juan Molinar Horcasitas, and Esperanza Palma; Peru: Cynthia McClintock, Charles Kenney, Felipe Ortiz de Zevallos, Carol Graham, and David Scott Palmer; Uruguay: David Altman, Rossana Castiglioni, and Juan Rial; Venezuela: Brian Crisp, José Molina Vega, David J. Myers, Juan Carlos Navarro, Juan Carlos Rey, and Luis Gómez Calcaño. I received additional advice from Steven Levitsky, Pierre Ostiguy, Donna Lee Van Cott. Unfortunately, no one supplied comprehensive feedback on my draft classification for Bolivia. Because the country specialists had no chance to respond to my attempts to reconcile their disagreements, they are in no way responsible for the final classification.

4. Votes for unclassified parties were excluded before calculating reliability because my inability to classify these parties was due to a lack of information about these parties rather than a lack of agreement among the experts.

5. The party could have a clear position on some dimension different from the ones reflected in my classification criteria; my coding criteria may not have been clear enough for some experts; the party might straddle a range of positions, making it unclear which one is the most representative; too few experts may have been consulted; some may have lacked relevant expertise; some may not have reviewed the draft diligently enough; or some might have supported the corrected classifications more frequently if they had been given the chance.

6. Obviously this indicator contains measurement error because there is some variation among parties within each bloc and there is no way to know whether the extremes are twice as far out as the center-right and center-left, or only 50 percent farther out, or three times as far out. However, I am persuaded that measurement with some error is better than no measurement at all, as long as the party classifications are valid.

7. For the purpose of measuring polarization in Argentina I have treated the Peronist-anti-Peronist confrontation as equivalent to an extreme difference in left-right terms before 1983 and a moderate difference from that year on.

8. It is important to remember that this is an indicator of left-right polarization only, and does not reflect the intense personal, ethnic, ins-outs, or other rivalries that sometimes exist between parties that are relatively close in left-right terms.

9. The figure for Western Europe comes from Bartolini and Mair 1990, p. 68. The figure for Latin America employs Bartolini and Mair's criteria for splits and regime changes to ensure comparability. If the criteria used for Table 3 are substituted, mean volatility for this set of elections is 26.6.

10. Mainwaring and Scully classified Mexico and Paraguay as ‘hegemonic systems in transition' rather than ‘institutionalized', although their ranking on the index is consistent with Sartori's classification. Also, Sartori classified Brazil as of 1966, at the beginning of the authoritarian regime, while Mainwaring and Scully discussed Brazil only during its democratic years (1945-64 and 1985-94). Finally, Mainwaring and Scully did not discuss four Central American countries that Sartori included--Guatemala, El Salvador, Honduras, and Nicaragua (all fluid).

11. The interval for Brazil excludes the 1966 and 1982 elections that marked the beginning and end of the authoritarian period. Technically, volatility was 100 in both of these elections, but because it was entirely artificial--the result of the military regime proscribing all previous parties by decree--for some purposes it would be misleading to average them in. Brazil's volatility is high enough without them.

12. The classified vote is simply the vote for parties that could be classified in this project. Rather recalculate the vote shares for all parties or blocs before calculating ENB, ENB_t was first calculated using percentages of the total valid vote and then multiplied by (1-U_t/100)2, which is mathematically equivalent. U_t is the percentage of the vote won by unclassified parties in election t.

Back to text