I define an equilibrium in a strategic environment with two entirely rational and sophisticated individuals, who form consistent assessment as classical statisticians based on their information (an equilibrium can be thought of as a steady state of some learning process). There are two equilibrium outcomes: (i) disagreement bears no idiosyncratic risks, minimizes aggregate welfare, individuals cannot recover the truth, and may hold different assessments; (ii) agreement is robust, maximizes welfare, and assessments coincide with the truth. A subjective Pareto criterion compares outcomes based on assessment sthat players may hold. Whereas agreement is Pareto efficient, disagreement subjectively Pareto-dominates agreement. Under equilibrium assessments, individuals disagree on redistribution. The model relates to “agreeing to dis-agree” (Aumann 1976) as well as trade and information (Milgrom and Stokey 1982).