Albert O. Hirschman, Exit, Voice, and Loyalty: Chapter 6, "On Spatial Duopoly and the Dynamics of Two-Party Systems"

By considering changes in quality not as absolute but as a trade off between preferences of different sets of consumers, Hirschman is able to analyze two-party political systems in terms of voice and exit.

In a monopolistic situation, where its share of the market can't be changed by exit, an organization will seek to produce a contested good at a quality that minimizes discontent.

In a situation where consumers at one end of the quality continuum have no substitute product, but the consumers at the other end are highly likely to defect to a substitute, voice is likely to be exercised only by the customers with nowhere to go. In this situation, profit (or market share) maximization would lead an organization to satisfy the volatile consumers no matter how great the discontent at the other end. But since minimizing discontent does come into play, a party will actually pursue a program at some distance from the desires of voters at the center of a two-party system.

Hotelling model of two-party system: parties will tend to move to the center, leaving voters on the outlying ends of the spectrum poorly served. Despite its pre-eminence, this thesis has not generally been borne out by the behavior of American political parties, which have positioned themselves at some distance on either side of the political center. One way of accounting for this failure lies in the model's neglect of elasticity of demand -- once a party moves far enough away from its base voters, it could begin to lose some of them. (To me, reduced turnout seems the most likely mechanism for this). Hirschman, however, does not find this criticism compelling, and offers a different explanation: the "captive" voters at the extreme ends of the continuum bring their parties closer to them by use of voice.