Electoral platform convergence is perceived unfavorably by both the popular press and many academic scholars. Arguably, "it does not provide enough choice" between candidates (Committee on Political Parties, 1950). This paper provides a formal account of the perceived negative effects of platform convergence. We show that when parties do not know voters' preferences precisely, all voters ex-ante prefer some platform divergence to convergence at the ex-ante median. After characterizing the unique symmetric equilibrium of competition between responsible (policy-motivated) parties, we conclude that all voters ex-ante prefer responsible parties to opportunistic (purely office-motivated) ones when parties are sufficiently ideologically polarized that platforms diverge, but not so polarized that they diverge excessively. However, greater polarization increases the scope for office benefits as an instrument for institutional design: We calculate the socially-optimal level of platform divergence and show that office benefits can be used to achieve this first-best outcome, if parties are sufficiently ideologically polarized.