Nobuhiro Hiwatari asked for some clarification concerning models that draw links from the electoral system to the party system. I think you have summarized Prof. Campbell's remarks concerning urbanization and its effects properly, and you are correct in pointing out that this demograhic change does indeed constitute a second variable -- sort of.

If you think of the "electoral system" as not just the allocation and voting rules, but also the districting scheme, then demograhic change such as urbanization (though exogenous) is still within the framework of the independent variable "electoral system" and you don't need a second variable.

However, even if you consider the environment a second variable (and this is perfectly acceptable) you miss the point of the claims about links between electoral rules and outcomes such as the number and type of parties. These models (including even Duverger's "Law") do not -- or should not -- ever claim that electoral rules by themselves determine anything. I refer you to almost anything Arend Lijphart has written about electoral laws and their political consequences (specifically a 1986 edited volume by him and Bernie Grofman with precisely that title -- Agathon Press). Think about a multiple regression analysis with, say, number of parties as the variable to be explained. All our models predict is that if you put electoral rules on the right-hand side, they should turn up significant in the directions normally discussed. These models do NOT claim that 100 percent of the variance in party systems will be explained, or that other explanatory variables -- such as ethnic or linguistic or religious or socioeconomic cleavages -- might not also matter.

PR, for example, by creating a low threshold, allows for a proliferation of parties -- it does not mandate it. Austria is a good example of PR with only two important parties. Germany arguably as well. Plurality systems, according to Duverger, tend toward 2-party competition, but this notion has been refined to specify that Duverger's mechanical and psychological effects may only work at the district level. Thus, if there is a strong regional ethnic group (like French-speakers in Quebec), the two parties that fight it out in that region might not be the same two parties as in some other region of the country (on this see Riker 1982 "The 2-party System and Duverger's Law..." APSR 76(4): 753-766).

So John Campbell's observation about the effects of urbanization is perfectly compatible with the idea that electoral rules will affect the number of parties in the system in predictable ways.

As for the question about the effects of electoral rules on the number of factions, the only thing I've read that discusses this is an article by Steve Reed, called something like "Extending Duverger's Law to Japan." Ask Steve for the correct citation or a summary. Nonetheless, I think it's important that these factions, however hard they fight over endorsements and votes and even policy, still saw fit for 40 years to remain under the same party umbrella. There is some cooperation there in addition to the obvious conflict, so the assumptions built into models of PARTY competition probably will not hold as well for factions. If they did, we should find every faction competing in every district, and the number of factions dropping to where that outcome could be sustained.

----------------------------------------------------------------
Michael F. Thies